Address review comments

build(cmake): fix missing Qt5::DBus link target for bundled linux package

CMakeLists.txt

@@ -363,7 +363,11 @@ if(ENABLE_QT)
 
         set(YUZU_QT_NO_CMAKE_SYSTEM_PATH "NO_CMAKE_SYSTEM_PATH")
     endif()
-    find_package(Qt5 ${QT_VERSION} REQUIRED COMPONENTS Widgets ${QT_PREFIX_HINT} ${YUZU_QT_NO_CMAKE_SYSTEM_PATH})
+    if ((${CMAKE_SYSTEM_NAME} STREQUAL "Linux") AND YUZU_USE_BUNDLED_QT)
+        find_package(Qt5 ${QT_VERSION} REQUIRED COMPONENTS Widgets DBus ${QT_PREFIX_HINT} ${YUZU_QT_NO_CMAKE_SYSTEM_PATH})
+    else()
+        find_package(Qt5 ${QT_VERSION} REQUIRED COMPONENTS Widgets ${QT_PREFIX_HINT} ${YUZU_QT_NO_CMAKE_SYSTEM_PATH})
+    endif()
     if (YUZU_USE_QT_WEB_ENGINE)
         find_package(Qt5 COMPONENTS WebEngineCore WebEngineWidgets)
     endif()

src/common/intrusive_red_black_tree.h

@@ -4,8 +4,6 @@
 
 #pragma once
 
-#include "common/alignment.h"
-#include "common/common_funcs.h"
 #include "common/parent_of_member.h"
 #include "common/tree.h"
 
@@ -17,33 +15,32 @@ class IntrusiveRedBlackTreeImpl;
 
 }
 
-#pragma pack(push, 4)
 struct IntrusiveRedBlackTreeNode {
-    YUZU_NON_COPYABLE(IntrusiveRedBlackTreeNode);
-
 public:
-    using RBEntry = freebsd::RBEntry<IntrusiveRedBlackTreeNode>;
+    using EntryType = RBEntry<IntrusiveRedBlackTreeNode>;
+
+    constexpr IntrusiveRedBlackTreeNode() = default;
+
+    void SetEntry(const EntryType& new_entry) {
+        entry = new_entry;
+    }
+
+    [[nodiscard]] EntryType& GetEntry() {
+        return entry;
+    }
+
+    [[nodiscard]] const EntryType& GetEntry() const {
+        return entry;
+    }
 
 private:
-    RBEntry m_entry;
+    EntryType entry{};
 
-public:
-    explicit IntrusiveRedBlackTreeNode() = default;
+    friend class impl::IntrusiveRedBlackTreeImpl;
 
-    [[nodiscard]] constexpr RBEntry& GetRBEntry() {
-        return m_entry;
-    }
-    [[nodiscard]] constexpr const RBEntry& GetRBEntry() const {
-        return m_entry;
-    }
-
-    constexpr void SetRBEntry(const RBEntry& entry) {
-        m_entry = entry;
-    }
+    template <class, class, class>
+    friend class IntrusiveRedBlackTree;
 };
-static_assert(sizeof(IntrusiveRedBlackTreeNode) ==
-              3 * sizeof(void*) + std::max<size_t>(sizeof(freebsd::RBColor), 4));
-#pragma pack(pop)
 
 template <class T, class Traits, class Comparator>
 class IntrusiveRedBlackTree;
@@ -51,17 +48,12 @@ class IntrusiveRedBlackTree;
 namespace impl {
 
 class IntrusiveRedBlackTreeImpl {
-    YUZU_NON_COPYABLE(IntrusiveRedBlackTreeImpl);
-
 private:
     template <class, class, class>
     friend class ::Common::IntrusiveRedBlackTree;
 
-private:
-    using RootType = freebsd::RBHead<IntrusiveRedBlackTreeNode>;
-
-private:
-    RootType m_root;
+    using RootType = RBHead<IntrusiveRedBlackTreeNode>;
+    RootType root;
 
 public:
     template <bool Const>
@@ -83,157 +75,155 @@ public:
         using iterator_category = std::bidirectional_iterator_tag;
         using value_type = typename IntrusiveRedBlackTreeImpl::value_type;
         using difference_type = typename IntrusiveRedBlackTreeImpl::difference_type;
-        using pointer = typename std::conditional<Const, IntrusiveRedBlackTreeImpl::const_pointer,
-                                                  IntrusiveRedBlackTreeImpl::pointer>::type;
-        using reference =
-            typename std::conditional<Const, IntrusiveRedBlackTreeImpl::const_reference,
-                                      IntrusiveRedBlackTreeImpl::reference>::type;
+        using pointer = std::conditional_t<Const, IntrusiveRedBlackTreeImpl::const_pointer,
+                                           IntrusiveRedBlackTreeImpl::pointer>;
+        using reference = std::conditional_t<Const, IntrusiveRedBlackTreeImpl::const_reference,
+                                             IntrusiveRedBlackTreeImpl::reference>;
 
     private:
-        pointer m_node;
+        pointer node;
 
     public:
-        constexpr explicit Iterator(pointer n) : m_node(n) {}
+        explicit Iterator(pointer n) : node(n) {}
 
-        constexpr bool operator==(const Iterator& rhs) const {
-            return m_node == rhs.m_node;
+        bool operator==(const Iterator& rhs) const {
+            return this->node == rhs.node;
         }
 
-        constexpr bool operator!=(const Iterator& rhs) const {
+        bool operator!=(const Iterator& rhs) const {
             return !(*this == rhs);
         }
 
-        constexpr pointer operator->() const {
-            return m_node;
+        pointer operator->() const {
+            return this->node;
         }
 
-        constexpr reference operator*() const {
-            return *m_node;
+        reference operator*() const {
+            return *this->node;
         }
 
-        constexpr Iterator& operator++() {
-            m_node = GetNext(m_node);
+        Iterator& operator++() {
+            this->node = GetNext(this->node);
             return *this;
         }
 
-        constexpr Iterator& operator--() {
-            m_node = GetPrev(m_node);
+        Iterator& operator--() {
+            this->node = GetPrev(this->node);
             return *this;
         }
 
-        constexpr Iterator operator++(int) {
+        Iterator operator++(int) {
             const Iterator it{*this};
             ++(*this);
             return it;
         }
 
-        constexpr Iterator operator--(int) {
+        Iterator operator--(int) {
             const Iterator it{*this};
             --(*this);
             return it;
         }
 
-        constexpr operator Iterator<true>() const {
-            return Iterator<true>(m_node);
+        operator Iterator<true>() const {
+            return Iterator<true>(this->node);
         }
     };
 
 private:
-    constexpr bool EmptyImpl() const {
-        return m_root.IsEmpty();
+    // Define accessors using RB_* functions.
+    bool EmptyImpl() const {
+        return root.IsEmpty();
     }
 
-    constexpr IntrusiveRedBlackTreeNode* GetMinImpl() const {
-        return freebsd::RB_MIN(const_cast<RootType&>(m_root));
+    IntrusiveRedBlackTreeNode* GetMinImpl() const {
+        return RB_MIN(const_cast<RootType*>(&root));
     }
 
-    constexpr IntrusiveRedBlackTreeNode* GetMaxImpl() const {
-        return freebsd::RB_MAX(const_cast<RootType&>(m_root));
+    IntrusiveRedBlackTreeNode* GetMaxImpl() const {
+        return RB_MAX(const_cast<RootType*>(&root));
     }
 
-    constexpr IntrusiveRedBlackTreeNode* RemoveImpl(IntrusiveRedBlackTreeNode* node) {
-        return freebsd::RB_REMOVE(m_root, node);
+    IntrusiveRedBlackTreeNode* RemoveImpl(IntrusiveRedBlackTreeNode* node) {
+        return RB_REMOVE(&root, node);
     }
 
 public:
-    static constexpr IntrusiveRedBlackTreeNode* GetNext(IntrusiveRedBlackTreeNode* node) {
-        return freebsd::RB_NEXT(node);
+    static IntrusiveRedBlackTreeNode* GetNext(IntrusiveRedBlackTreeNode* node) {
+        return RB_NEXT(node);
     }
 
-    static constexpr IntrusiveRedBlackTreeNode* GetPrev(IntrusiveRedBlackTreeNode* node) {
-        return freebsd::RB_PREV(node);
+    static IntrusiveRedBlackTreeNode* GetPrev(IntrusiveRedBlackTreeNode* node) {
+        return RB_PREV(node);
     }
 
-    static constexpr IntrusiveRedBlackTreeNode const* GetNext(
-        IntrusiveRedBlackTreeNode const* node) {
+    static const IntrusiveRedBlackTreeNode* GetNext(const IntrusiveRedBlackTreeNode* node) {
         return static_cast<const IntrusiveRedBlackTreeNode*>(
             GetNext(const_cast<IntrusiveRedBlackTreeNode*>(node)));
     }
 
-    static constexpr IntrusiveRedBlackTreeNode const* GetPrev(
-        IntrusiveRedBlackTreeNode const* node) {
+    static const IntrusiveRedBlackTreeNode* GetPrev(const IntrusiveRedBlackTreeNode* node) {
         return static_cast<const IntrusiveRedBlackTreeNode*>(
             GetPrev(const_cast<IntrusiveRedBlackTreeNode*>(node)));
     }
 
 public:
-    constexpr IntrusiveRedBlackTreeImpl() = default;
+    constexpr IntrusiveRedBlackTreeImpl() {}
 
     // Iterator accessors.
-    constexpr iterator begin() {
+    iterator begin() {
         return iterator(this->GetMinImpl());
     }
 
-    constexpr const_iterator begin() const {
+    const_iterator begin() const {
         return const_iterator(this->GetMinImpl());
     }
 
-    constexpr iterator end() {
+    iterator end() {
         return iterator(static_cast<IntrusiveRedBlackTreeNode*>(nullptr));
     }
 
-    constexpr const_iterator end() const {
+    const_iterator end() const {
         return const_iterator(static_cast<const IntrusiveRedBlackTreeNode*>(nullptr));
     }
 
-    constexpr const_iterator cbegin() const {
+    const_iterator cbegin() const {
         return this->begin();
     }
 
-    constexpr const_iterator cend() const {
+    const_iterator cend() const {
         return this->end();
     }
 
-    constexpr iterator iterator_to(reference ref) {
-        return iterator(std::addressof(ref));
+    iterator iterator_to(reference ref) {
+        return iterator(&ref);
     }
 
-    constexpr const_iterator iterator_to(const_reference ref) const {
-        return const_iterator(std::addressof(ref));
+    const_iterator iterator_to(const_reference ref) const {
+        return const_iterator(&ref);
     }
 
     // Content management.
-    constexpr bool empty() const {
+    bool empty() const {
         return this->EmptyImpl();
     }
 
-    constexpr reference back() {
+    reference back() {
         return *this->GetMaxImpl();
     }
 
-    constexpr const_reference back() const {
+    const_reference back() const {
         return *this->GetMaxImpl();
     }
 
-    constexpr reference front() {
+    reference front() {
         return *this->GetMinImpl();
     }
 
-    constexpr const_reference front() const {
+    const_reference front() const {
         return *this->GetMinImpl();
     }
 
-    constexpr iterator erase(iterator it) {
+    iterator erase(iterator it) {
         auto cur = std::addressof(*it);
         auto next = GetNext(cur);
         this->RemoveImpl(cur);
@@ -244,16 +234,16 @@ public:
 } // namespace impl
 
 template <typename T>
-concept HasRedBlackKeyType = requires {
-    { std::is_same<typename T::RedBlackKeyType, void>::value } -> std::convertible_to<bool>;
+concept HasLightCompareType = requires {
+    { std::is_same<typename T::LightCompareType, void>::value } -> std::convertible_to<bool>;
 };
 
 namespace impl {
 
     template <typename T, typename Default>
-    consteval auto* GetRedBlackKeyType() {
-        if constexpr (HasRedBlackKeyType<T>) {
-            return static_cast<typename T::RedBlackKeyType*>(nullptr);
+    consteval auto* GetLightCompareType() {
+        if constexpr (HasLightCompareType<T>) {
+            return static_cast<typename T::LightCompareType*>(nullptr);
         } else {
             return static_cast<Default*>(nullptr);
         }
@@ -262,18 +252,16 @@ namespace impl {
 } // namespace impl
 
 template <typename T, typename Default>
-using RedBlackKeyType =
-    typename std::remove_pointer<decltype(impl::GetRedBlackKeyType<T, Default>())>::type;
+using LightCompareType = std::remove_pointer_t<decltype(impl::GetLightCompareType<T, Default>())>;
 
 template <class T, class Traits, class Comparator>
 class IntrusiveRedBlackTree {
-    YUZU_NON_COPYABLE(IntrusiveRedBlackTree);
 
 public:
     using ImplType = impl::IntrusiveRedBlackTreeImpl;
 
 private:
-    ImplType m_impl;
+    ImplType impl{};
 
 public:
     template <bool Const>
@@ -289,9 +277,9 @@ public:
     using iterator = Iterator<false>;
     using const_iterator = Iterator<true>;
 
-    using key_type = RedBlackKeyType<Comparator, value_type>;
-    using const_key_pointer = const key_type*;
-    using const_key_reference = const key_type&;
+    using light_value_type = LightCompareType<Comparator, value_type>;
+    using const_light_pointer = const light_value_type*;
+    using const_light_reference = const light_value_type&;
 
     template <bool Const>
     class Iterator {
@@ -299,7 +287,7 @@ public:
         friend class IntrusiveRedBlackTree<T, Traits, Comparator>;
 
         using ImplIterator =
-            typename std::conditional<Const, ImplType::const_iterator, ImplType::iterator>::type;
+            std::conditional_t<Const, ImplType::const_iterator, ImplType::iterator>;
 
         using iterator_category = std::bidirectional_iterator_tag;
         using value_type = typename IntrusiveRedBlackTree::value_type;
@@ -310,201 +298,183 @@ public:
                                              IntrusiveRedBlackTree::reference>;
 
     private:
-        ImplIterator m_impl;
+        ImplIterator iterator;
 
     private:
-        constexpr explicit Iterator(ImplIterator it) : m_impl(it) {}
+        explicit Iterator(ImplIterator it) : iterator(it) {}
 
-        constexpr explicit Iterator(typename ImplIterator::pointer p) : m_impl(p) {}
+        explicit Iterator(typename std::conditional<Const, ImplType::const_iterator,
+                                                    ImplType::iterator>::type::pointer ptr)
+            : iterator(ptr) {}
 
-        constexpr ImplIterator GetImplIterator() const {
-            return m_impl;
+        ImplIterator GetImplIterator() const {
+            return this->iterator;
         }
 
     public:
-        constexpr bool operator==(const Iterator& rhs) const {
-            return m_impl == rhs.m_impl;
+        bool operator==(const Iterator& rhs) const {
+            return this->iterator == rhs.iterator;
         }
 
-        constexpr bool operator!=(const Iterator& rhs) const {
+        bool operator!=(const Iterator& rhs) const {
             return !(*this == rhs);
         }
 
-        constexpr pointer operator->() const {
-            return Traits::GetParent(std::addressof(*m_impl));
+        pointer operator->() const {
+            return Traits::GetParent(std::addressof(*this->iterator));
         }
 
-        constexpr reference operator*() const {
-            return *Traits::GetParent(std::addressof(*m_impl));
+        reference operator*() const {
+            return *Traits::GetParent(std::addressof(*this->iterator));
         }
 
-        constexpr Iterator& operator++() {
-            ++m_impl;
+        Iterator& operator++() {
+            ++this->iterator;
             return *this;
         }
 
-        constexpr Iterator& operator--() {
-            --m_impl;
+        Iterator& operator--() {
+            --this->iterator;
             return *this;
         }
 
-        constexpr Iterator operator++(int) {
+        Iterator operator++(int) {
             const Iterator it{*this};
-            ++m_impl;
+            ++this->iterator;
             return it;
         }
 
-        constexpr Iterator operator--(int) {
+        Iterator operator--(int) {
             const Iterator it{*this};
-            --m_impl;
+            --this->iterator;
             return it;
         }
 
-        constexpr operator Iterator<true>() const {
-            return Iterator<true>(m_impl);
+        operator Iterator<true>() const {
+            return Iterator<true>(this->iterator);
         }
     };
 
 private:
-    static constexpr int CompareImpl(const IntrusiveRedBlackTreeNode* lhs,
-                                     const IntrusiveRedBlackTreeNode* rhs) {
+    static int CompareImpl(const IntrusiveRedBlackTreeNode* lhs,
+                           const IntrusiveRedBlackTreeNode* rhs) {
         return Comparator::Compare(*Traits::GetParent(lhs), *Traits::GetParent(rhs));
     }
 
-    static constexpr int CompareKeyImpl(const_key_reference key,
-                                        const IntrusiveRedBlackTreeNode* rhs) {
-        return Comparator::Compare(key, *Traits::GetParent(rhs));
+    static int LightCompareImpl(const void* elm, const IntrusiveRedBlackTreeNode* rhs) {
+        return Comparator::Compare(*static_cast<const_light_pointer>(elm), *Traits::GetParent(rhs));
     }
 
     // Define accessors using RB_* functions.
-    constexpr IntrusiveRedBlackTreeNode* InsertImpl(IntrusiveRedBlackTreeNode* node) {
-        return freebsd::RB_INSERT(m_impl.m_root, node, CompareImpl);
+    IntrusiveRedBlackTreeNode* InsertImpl(IntrusiveRedBlackTreeNode* node) {
+        return RB_INSERT(&impl.root, node, CompareImpl);
     }
 
-    constexpr IntrusiveRedBlackTreeNode* FindImpl(IntrusiveRedBlackTreeNode const* node) const {
-        return freebsd::RB_FIND(const_cast<ImplType::RootType&>(m_impl.m_root),
-                                const_cast<IntrusiveRedBlackTreeNode*>(node), CompareImpl);
+    IntrusiveRedBlackTreeNode* FindImpl(const IntrusiveRedBlackTreeNode* node) const {
+        return RB_FIND(const_cast<ImplType::RootType*>(&impl.root),
+                       const_cast<IntrusiveRedBlackTreeNode*>(node), CompareImpl);
     }
 
-    constexpr IntrusiveRedBlackTreeNode* NFindImpl(IntrusiveRedBlackTreeNode const* node) const {
-        return freebsd::RB_NFIND(const_cast<ImplType::RootType&>(m_impl.m_root),
-                                 const_cast<IntrusiveRedBlackTreeNode*>(node), CompareImpl);
+    IntrusiveRedBlackTreeNode* NFindImpl(const IntrusiveRedBlackTreeNode* node) const {
+        return RB_NFIND(const_cast<ImplType::RootType*>(&impl.root),
+                        const_cast<IntrusiveRedBlackTreeNode*>(node), CompareImpl);
     }
 
-    constexpr IntrusiveRedBlackTreeNode* FindKeyImpl(const_key_reference key) const {
-        return freebsd::RB_FIND_KEY(const_cast<ImplType::RootType&>(m_impl.m_root), key,
-                                    CompareKeyImpl);
+    IntrusiveRedBlackTreeNode* FindLightImpl(const_light_pointer lelm) const {
+        return RB_FIND_LIGHT(const_cast<ImplType::RootType*>(&impl.root),
+                             static_cast<const void*>(lelm), LightCompareImpl);
     }
 
-    constexpr IntrusiveRedBlackTreeNode* NFindKeyImpl(const_key_reference key) const {
-        return freebsd::RB_NFIND_KEY(const_cast<ImplType::RootType&>(m_impl.m_root), key,
-                                     CompareKeyImpl);
-    }
-
-    constexpr IntrusiveRedBlackTreeNode* FindExistingImpl(
-        IntrusiveRedBlackTreeNode const* node) const {
-        return freebsd::RB_FIND_EXISTING(const_cast<ImplType::RootType&>(m_impl.m_root),
-                                         const_cast<IntrusiveRedBlackTreeNode*>(node), CompareImpl);
-    }
-
-    constexpr IntrusiveRedBlackTreeNode* FindExistingKeyImpl(const_key_reference key) const {
-        return freebsd::RB_FIND_EXISTING_KEY(const_cast<ImplType::RootType&>(m_impl.m_root), key,
-                                             CompareKeyImpl);
+    IntrusiveRedBlackTreeNode* NFindLightImpl(const_light_pointer lelm) const {
+        return RB_NFIND_LIGHT(const_cast<ImplType::RootType*>(&impl.root),
+                              static_cast<const void*>(lelm), LightCompareImpl);
     }
 
 public:
     constexpr IntrusiveRedBlackTree() = default;
 
     // Iterator accessors.
-    constexpr iterator begin() {
-        return iterator(m_impl.begin());
+    iterator begin() {
+        return iterator(this->impl.begin());
     }
 
-    constexpr const_iterator begin() const {
-        return const_iterator(m_impl.begin());
+    const_iterator begin() const {
+        return const_iterator(this->impl.begin());
     }
 
-    constexpr iterator end() {
-        return iterator(m_impl.end());
+    iterator end() {
+        return iterator(this->impl.end());
     }
 
-    constexpr const_iterator end() const {
-        return const_iterator(m_impl.end());
+    const_iterator end() const {
+        return const_iterator(this->impl.end());
     }
 
-    constexpr const_iterator cbegin() const {
+    const_iterator cbegin() const {
         return this->begin();
     }
 
-    constexpr const_iterator cend() const {
+    const_iterator cend() const {
         return this->end();
     }
 
-    constexpr iterator iterator_to(reference ref) {
-        return iterator(m_impl.iterator_to(*Traits::GetNode(std::addressof(ref))));
+    iterator iterator_to(reference ref) {
+        return iterator(this->impl.iterator_to(*Traits::GetNode(std::addressof(ref))));
     }
 
-    constexpr const_iterator iterator_to(const_reference ref) const {
-        return const_iterator(m_impl.iterator_to(*Traits::GetNode(std::addressof(ref))));
+    const_iterator iterator_to(const_reference ref) const {
+        return const_iterator(this->impl.iterator_to(*Traits::GetNode(std::addressof(ref))));
     }
 
     // Content management.
-    constexpr bool empty() const {
-        return m_impl.empty();
+    bool empty() const {
+        return this->impl.empty();
     }
 
-    constexpr reference back() {
-        return *Traits::GetParent(std::addressof(m_impl.back()));
+    reference back() {
+        return *Traits::GetParent(std::addressof(this->impl.back()));
     }
 
-    constexpr const_reference back() const {
-        return *Traits::GetParent(std::addressof(m_impl.back()));
+    const_reference back() const {
+        return *Traits::GetParent(std::addressof(this->impl.back()));
     }
 
-    constexpr reference front() {
-        return *Traits::GetParent(std::addressof(m_impl.front()));
+    reference front() {
+        return *Traits::GetParent(std::addressof(this->impl.front()));
     }
 
-    constexpr const_reference front() const {
-        return *Traits::GetParent(std::addressof(m_impl.front()));
+    const_reference front() const {
+        return *Traits::GetParent(std::addressof(this->impl.front()));
     }
 
-    constexpr iterator erase(iterator it) {
-        return iterator(m_impl.erase(it.GetImplIterator()));
+    iterator erase(iterator it) {
+        return iterator(this->impl.erase(it.GetImplIterator()));
     }
 
-    constexpr iterator insert(reference ref) {
+    iterator insert(reference ref) {
         ImplType::pointer node = Traits::GetNode(std::addressof(ref));
         this->InsertImpl(node);
         return iterator(node);
     }
 
-    constexpr iterator find(const_reference ref) const {
+    iterator find(const_reference ref) const {
         return iterator(this->FindImpl(Traits::GetNode(std::addressof(ref))));
     }
 
-    constexpr iterator nfind(const_reference ref) const {
+    iterator nfind(const_reference ref) const {
         return iterator(this->NFindImpl(Traits::GetNode(std::addressof(ref))));
     }
 
-    constexpr iterator find_key(const_key_reference ref) const {
-        return iterator(this->FindKeyImpl(ref));
+    iterator find_light(const_light_reference ref) const {
+        return iterator(this->FindLightImpl(std::addressof(ref)));
     }
 
-    constexpr iterator nfind_key(const_key_reference ref) const {
-        return iterator(this->NFindKeyImpl(ref));
-    }
-
-    constexpr iterator find_existing(const_reference ref) const {
-        return iterator(this->FindExistingImpl(Traits::GetNode(std::addressof(ref))));
-    }
-
-    constexpr iterator find_existing_key(const_key_reference ref) const {
-        return iterator(this->FindExistingKeyImpl(ref));
+    iterator nfind_light(const_light_reference ref) const {
+        return iterator(this->NFindLightImpl(std::addressof(ref)));
     }
 };
 
-template <auto T, class Derived = Common::impl::GetParentType<T>>
+template <auto T, class Derived = impl::GetParentType<T>>
 class IntrusiveRedBlackTreeMemberTraits;
 
 template <class Parent, IntrusiveRedBlackTreeNode Parent::*Member, class Derived>
@@ -528,16 +498,19 @@ private:
         return std::addressof(parent->*Member);
     }
 
-    static Derived* GetParent(IntrusiveRedBlackTreeNode* node) {
-        return Common::GetParentPointer<Member, Derived>(node);
+    static constexpr Derived* GetParent(IntrusiveRedBlackTreeNode* node) {
+        return GetParentPointer<Member, Derived>(node);
     }
 
-    static Derived const* GetParent(IntrusiveRedBlackTreeNode const* node) {
-        return Common::GetParentPointer<Member, Derived>(node);
+    static constexpr Derived const* GetParent(const IntrusiveRedBlackTreeNode* node) {
+        return GetParentPointer<Member, Derived>(node);
     }
+
+private:
+    static constexpr TypedStorage<Derived> DerivedStorage = {};
 };
 
-template <auto T, class Derived = Common::impl::GetParentType<T>>
+template <auto T, class Derived = impl::GetParentType<T>>
 class IntrusiveRedBlackTreeMemberTraitsDeferredAssert;
 
 template <class Parent, IntrusiveRedBlackTreeNode Parent::*Member, class Derived>
@@ -548,6 +521,11 @@ public:
         IntrusiveRedBlackTree<Derived, IntrusiveRedBlackTreeMemberTraitsDeferredAssert, Comparator>;
     using TreeTypeImpl = impl::IntrusiveRedBlackTreeImpl;
 
+    static constexpr bool IsValid() {
+        TypedStorage<Derived> DerivedStorage = {};
+        return GetParent(GetNode(GetPointer(DerivedStorage))) == GetPointer(DerivedStorage);
+    }
+
 private:
     template <class, class, class>
     friend class IntrusiveRedBlackTree;
@@ -562,36 +540,30 @@ private:
         return std::addressof(parent->*Member);
     }
 
-    static Derived* GetParent(IntrusiveRedBlackTreeNode* node) {
-        return Common::GetParentPointer<Member, Derived>(node);
+    static constexpr Derived* GetParent(IntrusiveRedBlackTreeNode* node) {
+        return GetParentPointer<Member, Derived>(node);
     }
 
-    static Derived const* GetParent(IntrusiveRedBlackTreeNode const* node) {
-        return Common::GetParentPointer<Member, Derived>(node);
+    static constexpr Derived const* GetParent(const IntrusiveRedBlackTreeNode* node) {
+        return GetParentPointer<Member, Derived>(node);
     }
 };
 
 template <class Derived>
-class alignas(void*) IntrusiveRedBlackTreeBaseNode : public IntrusiveRedBlackTreeNode {
+class IntrusiveRedBlackTreeBaseNode : public IntrusiveRedBlackTreeNode {
 public:
-    using IntrusiveRedBlackTreeNode::IntrusiveRedBlackTreeNode;
-
     constexpr Derived* GetPrev() {
-        return static_cast<Derived*>(static_cast<IntrusiveRedBlackTreeBaseNode*>(
-            impl::IntrusiveRedBlackTreeImpl::GetPrev(this)));
+        return static_cast<Derived*>(impl::IntrusiveRedBlackTreeImpl::GetPrev(this));
     }
     constexpr const Derived* GetPrev() const {
-        return static_cast<const Derived*>(static_cast<const IntrusiveRedBlackTreeBaseNode*>(
-            impl::IntrusiveRedBlackTreeImpl::GetPrev(this)));
+        return static_cast<const Derived*>(impl::IntrusiveRedBlackTreeImpl::GetPrev(this));
     }
 
     constexpr Derived* GetNext() {
-        return static_cast<Derived*>(static_cast<IntrusiveRedBlackTreeBaseNode*>(
-            impl::IntrusiveRedBlackTreeImpl::GetNext(this)));
+        return static_cast<Derived*>(impl::IntrusiveRedBlackTreeImpl::GetNext(this));
     }
     constexpr const Derived* GetNext() const {
-        return static_cast<const Derived*>(static_cast<const IntrusiveRedBlackTreeBaseNode*>(
-            impl::IntrusiveRedBlackTreeImpl::GetNext(this)));
+        return static_cast<const Derived*>(impl::IntrusiveRedBlackTreeImpl::GetNext(this));
     }
 };
 
@@ -609,22 +581,19 @@ private:
     friend class impl::IntrusiveRedBlackTreeImpl;
 
     static constexpr IntrusiveRedBlackTreeNode* GetNode(Derived* parent) {
-        return static_cast<IntrusiveRedBlackTreeNode*>(
-            static_cast<IntrusiveRedBlackTreeBaseNode<Derived>*>(parent));
+        return static_cast<IntrusiveRedBlackTreeNode*>(parent);
     }
 
     static constexpr IntrusiveRedBlackTreeNode const* GetNode(Derived const* parent) {
-        return static_cast<const IntrusiveRedBlackTreeNode*>(
-            static_cast<const IntrusiveRedBlackTreeBaseNode<Derived>*>(parent));
+        return static_cast<const IntrusiveRedBlackTreeNode*>(parent);
     }
 
     static constexpr Derived* GetParent(IntrusiveRedBlackTreeNode* node) {
-        return static_cast<Derived*>(static_cast<IntrusiveRedBlackTreeBaseNode<Derived>*>(node));
+        return static_cast<Derived*>(node);
     }
 
-    static constexpr Derived const* GetParent(IntrusiveRedBlackTreeNode const* node) {
-        return static_cast<const Derived*>(
-            static_cast<const IntrusiveRedBlackTreeBaseNode<Derived>*>(node));
+    static constexpr Derived const* GetParent(const IntrusiveRedBlackTreeNode* node) {
+        return static_cast<const Derived*>(node);
     }
 };
 

src/common/tree.h

@@ -29,8 +29,6 @@
 
 #pragma once
 
-#include "common/assert.h"
-
 /*
  * This file defines data structures for red-black trees.
  *
@@ -45,445 +43,246 @@
  * The maximum height of a red-black tree is 2lg (n+1).
  */
 
-namespace Common::freebsd {
+#include "common/assert.h"
 
-enum class RBColor {
-    RB_BLACK = 0,
-    RB_RED = 1,
+namespace Common {
+template <typename T>
+class RBHead {
+public:
+    [[nodiscard]] T* Root() {
+        return rbh_root;
+    }
+
+    [[nodiscard]] const T* Root() const {
+        return rbh_root;
+    }
+
+    void SetRoot(T* root) {
+        rbh_root = root;
+    }
+
+    [[nodiscard]] bool IsEmpty() const {
+        return Root() == nullptr;
+    }
+
+private:
+    T* rbh_root = nullptr;
+};
+
+enum class EntryColor {
+    Black,
+    Red,
 };
 
-#pragma pack(push, 4)
 template <typename T>
 class RBEntry {
-private:
-    T* m_rbe_left{};
-    T* m_rbe_right{};
-    T* m_rbe_parent{};
-    RBColor m_rbe_color{RBColor::RB_BLACK};
-
 public:
-    constexpr RBEntry() = default;
-
-    [[nodiscard]] constexpr T* Left() {
-        return m_rbe_left;
-    }
-    [[nodiscard]] constexpr const T* Left() const {
-        return m_rbe_left;
+    [[nodiscard]] T* Left() {
+        return rbe_left;
     }
 
-    constexpr void SetLeft(T* e) {
-        m_rbe_left = e;
+    [[nodiscard]] const T* Left() const {
+        return rbe_left;
     }
 
-    [[nodiscard]] constexpr T* Right() {
-        return m_rbe_right;
-    }
-    [[nodiscard]] constexpr const T* Right() const {
-        return m_rbe_right;
+    void SetLeft(T* left) {
+        rbe_left = left;
     }
 
-    constexpr void SetRight(T* e) {
-        m_rbe_right = e;
+    [[nodiscard]] T* Right() {
+        return rbe_right;
     }
 
-    [[nodiscard]] constexpr T* Parent() {
-        return m_rbe_parent;
-    }
-    [[nodiscard]] constexpr const T* Parent() const {
-        return m_rbe_parent;
+    [[nodiscard]] const T* Right() const {
+        return rbe_right;
     }
 
-    constexpr void SetParent(T* e) {
-        m_rbe_parent = e;
+    void SetRight(T* right) {
+        rbe_right = right;
     }
 
-    [[nodiscard]] constexpr bool IsBlack() const {
-        return m_rbe_color == RBColor::RB_BLACK;
-    }
-    [[nodiscard]] constexpr bool IsRed() const {
-        return m_rbe_color == RBColor::RB_RED;
-    }
-    [[nodiscard]] constexpr RBColor Color() const {
-        return m_rbe_color;
+    [[nodiscard]] T* Parent() {
+        return rbe_parent;
     }
 
-    constexpr void SetColor(RBColor c) {
-        m_rbe_color = c;
+    [[nodiscard]] const T* Parent() const {
+        return rbe_parent;
     }
-};
-#pragma pack(pop)
 
-template <typename T>
-struct CheckRBEntry {
-    static constexpr bool value = false;
-};
-template <typename T>
-struct CheckRBEntry<RBEntry<T>> {
-    static constexpr bool value = true;
-};
+    void SetParent(T* parent) {
+        rbe_parent = parent;
+    }
 
-template <typename T>
-concept IsRBEntry = CheckRBEntry<T>::value;
+    [[nodiscard]] bool IsBlack() const {
+        return rbe_color == EntryColor::Black;
+    }
 
-template <typename T>
-concept HasRBEntry = requires(T& t, const T& ct) {
-    { t.GetRBEntry() } -> std::same_as<RBEntry<T>&>;
-    { ct.GetRBEntry() } -> std::same_as<const RBEntry<T>&>;
-};
+    [[nodiscard]] bool IsRed() const {
+        return rbe_color == EntryColor::Red;
+    }
+
+    [[nodiscard]] EntryColor Color() const {
+        return rbe_color;
+    }
+
+    void SetColor(EntryColor color) {
+        rbe_color = color;
+    }
 
-template <typename T>
-requires HasRBEntry<T>
-class RBHead {
 private:
-    T* m_rbh_root = nullptr;
-
-public:
-    [[nodiscard]] constexpr T* Root() {
-        return m_rbh_root;
-    }
-    [[nodiscard]] constexpr const T* Root() const {
-        return m_rbh_root;
-    }
-    constexpr void SetRoot(T* root) {
-        m_rbh_root = root;
-    }
-
-    [[nodiscard]] constexpr bool IsEmpty() const {
-        return this->Root() == nullptr;
-    }
+    T* rbe_left = nullptr;
+    T* rbe_right = nullptr;
+    T* rbe_parent = nullptr;
+    EntryColor rbe_color{};
 };
 
-template <typename T>
-requires HasRBEntry<T>
-[[nodiscard]] constexpr RBEntry<T>& RB_ENTRY(T* t) {
-    return t->GetRBEntry();
-}
-template <typename T>
-requires HasRBEntry<T>
-[[nodiscard]] constexpr const RBEntry<T>& RB_ENTRY(const T* t) {
-    return t->GetRBEntry();
+template <typename Node>
+[[nodiscard]] RBEntry<Node>& RB_ENTRY(Node* node) {
+    return node->GetEntry();
 }
 
-template <typename T>
-requires HasRBEntry<T>
-[[nodiscard]] constexpr T* RB_LEFT(T* t) {
-    return RB_ENTRY(t).Left();
-}
-template <typename T>
-requires HasRBEntry<T>
-[[nodiscard]] constexpr const T* RB_LEFT(const T* t) {
-    return RB_ENTRY(t).Left();
+template <typename Node>
+[[nodiscard]] const RBEntry<Node>& RB_ENTRY(const Node* node) {
+    return node->GetEntry();
 }
 
-template <typename T>
-requires HasRBEntry<T>
-[[nodiscard]] constexpr T* RB_RIGHT(T* t) {
-    return RB_ENTRY(t).Right();
-}
-template <typename T>
-requires HasRBEntry<T>
-[[nodiscard]] constexpr const T* RB_RIGHT(const T* t) {
-    return RB_ENTRY(t).Right();
+template <typename Node>
+[[nodiscard]] Node* RB_PARENT(Node* node) {
+    return RB_ENTRY(node).Parent();
 }
 
-template <typename T>
-requires HasRBEntry<T>
-[[nodiscard]] constexpr T* RB_PARENT(T* t) {
-    return RB_ENTRY(t).Parent();
-}
-template <typename T>
-requires HasRBEntry<T>
-[[nodiscard]] constexpr const T* RB_PARENT(const T* t) {
-    return RB_ENTRY(t).Parent();
+template <typename Node>
+[[nodiscard]] const Node* RB_PARENT(const Node* node) {
+    return RB_ENTRY(node).Parent();
 }
 
-template <typename T>
-requires HasRBEntry<T>
-constexpr void RB_SET_LEFT(T* t, T* e) {
-    RB_ENTRY(t).SetLeft(e);
-}
-template <typename T>
-requires HasRBEntry<T>
-constexpr void RB_SET_RIGHT(T* t, T* e) {
-    RB_ENTRY(t).SetRight(e);
-}
-template <typename T>
-requires HasRBEntry<T>
-constexpr void RB_SET_PARENT(T* t, T* e) {
-    RB_ENTRY(t).SetParent(e);
+template <typename Node>
+void RB_SET_PARENT(Node* node, Node* parent) {
+    return RB_ENTRY(node).SetParent(parent);
 }
 
-template <typename T>
-requires HasRBEntry<T>
-[[nodiscard]] constexpr bool RB_IS_BLACK(const T* t) {
-    return RB_ENTRY(t).IsBlack();
-}
-template <typename T>
-requires HasRBEntry<T>
-[[nodiscard]] constexpr bool RB_IS_RED(const T* t) {
-    return RB_ENTRY(t).IsRed();
+template <typename Node>
+[[nodiscard]] Node* RB_LEFT(Node* node) {
+    return RB_ENTRY(node).Left();
 }
 
-template <typename T>
-requires HasRBEntry<T>
-[[nodiscard]] constexpr RBColor RB_COLOR(const T* t) {
-    return RB_ENTRY(t).Color();
+template <typename Node>
+[[nodiscard]] const Node* RB_LEFT(const Node* node) {
+    return RB_ENTRY(node).Left();
 }
 
-template <typename T>
-requires HasRBEntry<T>
-constexpr void RB_SET_COLOR(T* t, RBColor c) {
-    RB_ENTRY(t).SetColor(c);
+template <typename Node>
+void RB_SET_LEFT(Node* node, Node* left) {
+    return RB_ENTRY(node).SetLeft(left);
 }
 
-template <typename T>
-requires HasRBEntry<T>
-constexpr void RB_SET(T* elm, T* parent) {
-    auto& rb_entry = RB_ENTRY(elm);
-    rb_entry.SetParent(parent);
-    rb_entry.SetLeft(nullptr);
-    rb_entry.SetRight(nullptr);
-    rb_entry.SetColor(RBColor::RB_RED);
+template <typename Node>
+[[nodiscard]] Node* RB_RIGHT(Node* node) {
+    return RB_ENTRY(node).Right();
 }
 
-template <typename T>
-requires HasRBEntry<T>
-constexpr void RB_SET_BLACKRED(T* black, T* red) {
-    RB_SET_COLOR(black, RBColor::RB_BLACK);
-    RB_SET_COLOR(red, RBColor::RB_RED);
+template <typename Node>
+[[nodiscard]] const Node* RB_RIGHT(const Node* node) {
+    return RB_ENTRY(node).Right();
 }
 
-template <typename T>
-requires HasRBEntry<T>
-constexpr void RB_ROTATE_LEFT(RBHead<T>& head, T* elm, T*& tmp) {
+template <typename Node>
+void RB_SET_RIGHT(Node* node, Node* right) {
+    return RB_ENTRY(node).SetRight(right);
+}
+
+template <typename Node>
+[[nodiscard]] bool RB_IS_BLACK(const Node* node) {
+    return RB_ENTRY(node).IsBlack();
+}
+
+template <typename Node>
+[[nodiscard]] bool RB_IS_RED(const Node* node) {
+    return RB_ENTRY(node).IsRed();
+}
+
+template <typename Node>
+[[nodiscard]] EntryColor RB_COLOR(const Node* node) {
+    return RB_ENTRY(node).Color();
+}
+
+template <typename Node>
+void RB_SET_COLOR(Node* node, EntryColor color) {
+    return RB_ENTRY(node).SetColor(color);
+}
+
+template <typename Node>
+void RB_SET(Node* node, Node* parent) {
+    auto& entry = RB_ENTRY(node);
+    entry.SetParent(parent);
+    entry.SetLeft(nullptr);
+    entry.SetRight(nullptr);
+    entry.SetColor(EntryColor::Red);
+}
+
+template <typename Node>
+void RB_SET_BLACKRED(Node* black, Node* red) {
+    RB_SET_COLOR(black, EntryColor::Black);
+    RB_SET_COLOR(red, EntryColor::Red);
+}
+
+template <typename Node>
+void RB_ROTATE_LEFT(RBHead<Node>* head, Node* elm, Node*& tmp) {
     tmp = RB_RIGHT(elm);
-    if (RB_SET_RIGHT(elm, RB_LEFT(tmp)); RB_RIGHT(elm) != nullptr) {
+    RB_SET_RIGHT(elm, RB_LEFT(tmp));
+    if (RB_RIGHT(elm) != nullptr) {
         RB_SET_PARENT(RB_LEFT(tmp), elm);
     }
 
-    if (RB_SET_PARENT(tmp, RB_PARENT(elm)); RB_PARENT(tmp) != nullptr) {
+    RB_SET_PARENT(tmp, RB_PARENT(elm));
+    if (RB_PARENT(tmp) != nullptr) {
         if (elm == RB_LEFT(RB_PARENT(elm))) {
             RB_SET_LEFT(RB_PARENT(elm), tmp);
         } else {
             RB_SET_RIGHT(RB_PARENT(elm), tmp);
         }
     } else {
-        head.SetRoot(tmp);
+        head->SetRoot(tmp);
     }
 
     RB_SET_LEFT(tmp, elm);
     RB_SET_PARENT(elm, tmp);
 }
 
-template <typename T>
-requires HasRBEntry<T>
-constexpr void RB_ROTATE_RIGHT(RBHead<T>& head, T* elm, T*& tmp) {
+template <typename Node>
+void RB_ROTATE_RIGHT(RBHead<Node>* head, Node* elm, Node*& tmp) {
     tmp = RB_LEFT(elm);
-    if (RB_SET_LEFT(elm, RB_RIGHT(tmp)); RB_LEFT(elm) != nullptr) {
+    RB_SET_LEFT(elm, RB_RIGHT(tmp));
+    if (RB_LEFT(elm) != nullptr) {
         RB_SET_PARENT(RB_RIGHT(tmp), elm);
     }
 
-    if (RB_SET_PARENT(tmp, RB_PARENT(elm)); RB_PARENT(tmp) != nullptr) {
+    RB_SET_PARENT(tmp, RB_PARENT(elm));
+    if (RB_PARENT(tmp) != nullptr) {
         if (elm == RB_LEFT(RB_PARENT(elm))) {
             RB_SET_LEFT(RB_PARENT(elm), tmp);
         } else {
             RB_SET_RIGHT(RB_PARENT(elm), tmp);
         }
     } else {
-        head.SetRoot(tmp);
+        head->SetRoot(tmp);
     }
 
     RB_SET_RIGHT(tmp, elm);
     RB_SET_PARENT(elm, tmp);
 }
 
-template <typename T>
-requires HasRBEntry<T>
-constexpr void RB_REMOVE_COLOR(RBHead<T>& head, T* parent, T* elm) {
-    T* tmp;
-    while ((elm == nullptr || RB_IS_BLACK(elm)) && elm != head.Root()) {
-        if (RB_LEFT(parent) == elm) {
-            tmp = RB_RIGHT(parent);
-            if (RB_IS_RED(tmp)) {
-                RB_SET_BLACKRED(tmp, parent);
-                RB_ROTATE_LEFT(head, parent, tmp);
-                tmp = RB_RIGHT(parent);
-            }
+template <typename Node>
+void RB_INSERT_COLOR(RBHead<Node>* head, Node* elm) {
+    Node* parent = nullptr;
+    Node* tmp = nullptr;
 
-            if ((RB_LEFT(tmp) == nullptr || RB_IS_BLACK(RB_LEFT(tmp))) &&
-                (RB_RIGHT(tmp) == nullptr || RB_IS_BLACK(RB_RIGHT(tmp)))) {
-                RB_SET_COLOR(tmp, RBColor::RB_RED);
-                elm = parent;
-                parent = RB_PARENT(elm);
-            } else {
-                if (RB_RIGHT(tmp) == nullptr || RB_IS_BLACK(RB_RIGHT(tmp))) {
-                    T* oleft;
-                    if ((oleft = RB_LEFT(tmp)) != nullptr) {
-                        RB_SET_COLOR(oleft, RBColor::RB_BLACK);
-                    }
-
-                    RB_SET_COLOR(tmp, RBColor::RB_RED);
-                    RB_ROTATE_RIGHT(head, tmp, oleft);
-                    tmp = RB_RIGHT(parent);
-                }
-
-                RB_SET_COLOR(tmp, RB_COLOR(parent));
-                RB_SET_COLOR(parent, RBColor::RB_BLACK);
-                if (RB_RIGHT(tmp)) {
-                    RB_SET_COLOR(RB_RIGHT(tmp), RBColor::RB_BLACK);
-                }
-
-                RB_ROTATE_LEFT(head, parent, tmp);
-                elm = head.Root();
-                break;
-            }
-        } else {
-            tmp = RB_LEFT(parent);
-            if (RB_IS_RED(tmp)) {
-                RB_SET_BLACKRED(tmp, parent);
-                RB_ROTATE_RIGHT(head, parent, tmp);
-                tmp = RB_LEFT(parent);
-            }
-
-            if ((RB_LEFT(tmp) == nullptr || RB_IS_BLACK(RB_LEFT(tmp))) &&
-                (RB_RIGHT(tmp) == nullptr || RB_IS_BLACK(RB_RIGHT(tmp)))) {
-                RB_SET_COLOR(tmp, RBColor::RB_RED);
-                elm = parent;
-                parent = RB_PARENT(elm);
-            } else {
-                if (RB_LEFT(tmp) == nullptr || RB_IS_BLACK(RB_LEFT(tmp))) {
-                    T* oright;
-                    if ((oright = RB_RIGHT(tmp)) != nullptr) {
-                        RB_SET_COLOR(oright, RBColor::RB_BLACK);
-                    }
-
-                    RB_SET_COLOR(tmp, RBColor::RB_RED);
-                    RB_ROTATE_LEFT(head, tmp, oright);
-                    tmp = RB_LEFT(parent);
-                }
-
-                RB_SET_COLOR(tmp, RB_COLOR(parent));
-                RB_SET_COLOR(parent, RBColor::RB_BLACK);
-
-                if (RB_LEFT(tmp)) {
-                    RB_SET_COLOR(RB_LEFT(tmp), RBColor::RB_BLACK);
-                }
-
-                RB_ROTATE_RIGHT(head, parent, tmp);
-                elm = head.Root();
-                break;
-            }
-        }
-    }
-
-    if (elm) {
-        RB_SET_COLOR(elm, RBColor::RB_BLACK);
-    }
-}
-
-template <typename T>
-requires HasRBEntry<T>
-constexpr T* RB_REMOVE(RBHead<T>& head, T* elm) {
-    T* child = nullptr;
-    T* parent = nullptr;
-    T* old = elm;
-    RBColor color = RBColor::RB_BLACK;
-
-    if (RB_LEFT(elm) == nullptr) {
-        child = RB_RIGHT(elm);
-    } else if (RB_RIGHT(elm) == nullptr) {
-        child = RB_LEFT(elm);
-    } else {
-        T* left;
-        elm = RB_RIGHT(elm);
-        while ((left = RB_LEFT(elm)) != nullptr) {
-            elm = left;
-        }
-
-        child = RB_RIGHT(elm);
-        parent = RB_PARENT(elm);
-        color = RB_COLOR(elm);
-
-        if (child) {
-            RB_SET_PARENT(child, parent);
-        }
-
-        if (parent) {
-            if (RB_LEFT(parent) == elm) {
-                RB_SET_LEFT(parent, child);
-            } else {
-                RB_SET_RIGHT(parent, child);
-            }
-        } else {
-            head.SetRoot(child);
-        }
-
-        if (RB_PARENT(elm) == old) {
-            parent = elm;
-        }
-
-        elm->SetRBEntry(old->GetRBEntry());
-
-        if (RB_PARENT(old)) {
-            if (RB_LEFT(RB_PARENT(old)) == old) {
-                RB_SET_LEFT(RB_PARENT(old), elm);
-            } else {
-                RB_SET_RIGHT(RB_PARENT(old), elm);
-            }
-        } else {
-            head.SetRoot(elm);
-        }
-
-        RB_SET_PARENT(RB_LEFT(old), elm);
-
-        if (RB_RIGHT(old)) {
-            RB_SET_PARENT(RB_RIGHT(old), elm);
-        }
-
-        if (parent) {
-            left = parent;
-        }
-
-        if (color == RBColor::RB_BLACK) {
-            RB_REMOVE_COLOR(head, parent, child);
-        }
-
-        return old;
-    }
-
-    parent = RB_PARENT(elm);
-    color = RB_COLOR(elm);
-
-    if (child) {
-        RB_SET_PARENT(child, parent);
-    }
-    if (parent) {
-        if (RB_LEFT(parent) == elm) {
-            RB_SET_LEFT(parent, child);
-        } else {
-            RB_SET_RIGHT(parent, child);
-        }
-    } else {
-        head.SetRoot(child);
-    }
-
-    if (color == RBColor::RB_BLACK) {
-        RB_REMOVE_COLOR(head, parent, child);
-    }
-
-    return old;
-}
-
-template <typename T>
-requires HasRBEntry<T>
-constexpr void RB_INSERT_COLOR(RBHead<T>& head, T* elm) {
-    T *parent = nullptr, *tmp = nullptr;
     while ((parent = RB_PARENT(elm)) != nullptr && RB_IS_RED(parent)) {
-        T* gparent = RB_PARENT(parent);
+        Node* gparent = RB_PARENT(parent);
         if (parent == RB_LEFT(gparent)) {
             tmp = RB_RIGHT(gparent);
             if (tmp && RB_IS_RED(tmp)) {
-                RB_SET_COLOR(tmp, RBColor::RB_BLACK);
+                RB_SET_COLOR(tmp, EntryColor::Black);
                 RB_SET_BLACKRED(parent, gparent);
                 elm = gparent;
                 continue;
@@ -501,7 +300,7 @@ constexpr void RB_INSERT_COLOR(RBHead<T>& head, T* elm) {
         } else {
             tmp = RB_LEFT(gparent);
             if (tmp && RB_IS_RED(tmp)) {
-                RB_SET_COLOR(tmp, RBColor::RB_BLACK);
+                RB_SET_COLOR(tmp, EntryColor::Black);
                 RB_SET_BLACKRED(parent, gparent);
                 elm = gparent;
                 continue;
@@ -519,14 +318,194 @@ constexpr void RB_INSERT_COLOR(RBHead<T>& head, T* elm) {
         }
     }
 
-    RB_SET_COLOR(head.Root(), RBColor::RB_BLACK);
+    RB_SET_COLOR(head->Root(), EntryColor::Black);
 }
 
-template <typename T, typename Compare>
-requires HasRBEntry<T>
-constexpr T* RB_INSERT(RBHead<T>& head, T* elm, Compare cmp) {
-    T* parent = nullptr;
-    T* tmp = head.Root();
+template <typename Node>
+void RB_REMOVE_COLOR(RBHead<Node>* head, Node* parent, Node* elm) {
+    Node* tmp;
+    while ((elm == nullptr || RB_IS_BLACK(elm)) && elm != head->Root() && parent != nullptr) {
+        if (RB_LEFT(parent) == elm) {
+            tmp = RB_RIGHT(parent);
+            if (!tmp) {
+                ASSERT_MSG(false, "tmp is invalid!");
+                break;
+            }
+            if (RB_IS_RED(tmp)) {
+                RB_SET_BLACKRED(tmp, parent);
+                RB_ROTATE_LEFT(head, parent, tmp);
+                tmp = RB_RIGHT(parent);
+            }
+
+            if ((RB_LEFT(tmp) == nullptr || RB_IS_BLACK(RB_LEFT(tmp))) &&
+                (RB_RIGHT(tmp) == nullptr || RB_IS_BLACK(RB_RIGHT(tmp)))) {
+                RB_SET_COLOR(tmp, EntryColor::Red);
+                elm = parent;
+                parent = RB_PARENT(elm);
+            } else {
+                if (RB_RIGHT(tmp) == nullptr || RB_IS_BLACK(RB_RIGHT(tmp))) {
+                    Node* oleft;
+                    if ((oleft = RB_LEFT(tmp)) != nullptr) {
+                        RB_SET_COLOR(oleft, EntryColor::Black);
+                    }
+
+                    RB_SET_COLOR(tmp, EntryColor::Red);
+                    RB_ROTATE_RIGHT(head, tmp, oleft);
+                    tmp = RB_RIGHT(parent);
+                }
+
+                RB_SET_COLOR(tmp, RB_COLOR(parent));
+                RB_SET_COLOR(parent, EntryColor::Black);
+                if (RB_RIGHT(tmp)) {
+                    RB_SET_COLOR(RB_RIGHT(tmp), EntryColor::Black);
+                }
+
+                RB_ROTATE_LEFT(head, parent, tmp);
+                elm = head->Root();
+                break;
+            }
+        } else {
+            tmp = RB_LEFT(parent);
+            if (RB_IS_RED(tmp)) {
+                RB_SET_BLACKRED(tmp, parent);
+                RB_ROTATE_RIGHT(head, parent, tmp);
+                tmp = RB_LEFT(parent);
+            }
+
+            if (!tmp) {
+                ASSERT_MSG(false, "tmp is invalid!");
+                break;
+            }
+
+            if ((RB_LEFT(tmp) == nullptr || RB_IS_BLACK(RB_LEFT(tmp))) &&
+                (RB_RIGHT(tmp) == nullptr || RB_IS_BLACK(RB_RIGHT(tmp)))) {
+                RB_SET_COLOR(tmp, EntryColor::Red);
+                elm = parent;
+                parent = RB_PARENT(elm);
+            } else {
+                if (RB_LEFT(tmp) == nullptr || RB_IS_BLACK(RB_LEFT(tmp))) {
+                    Node* oright;
+                    if ((oright = RB_RIGHT(tmp)) != nullptr) {
+                        RB_SET_COLOR(oright, EntryColor::Black);
+                    }
+
+                    RB_SET_COLOR(tmp, EntryColor::Red);
+                    RB_ROTATE_LEFT(head, tmp, oright);
+                    tmp = RB_LEFT(parent);
+                }
+
+                RB_SET_COLOR(tmp, RB_COLOR(parent));
+                RB_SET_COLOR(parent, EntryColor::Black);
+
+                if (RB_LEFT(tmp)) {
+                    RB_SET_COLOR(RB_LEFT(tmp), EntryColor::Black);
+                }
+
+                RB_ROTATE_RIGHT(head, parent, tmp);
+                elm = head->Root();
+                break;
+            }
+        }
+    }
+
+    if (elm) {
+        RB_SET_COLOR(elm, EntryColor::Black);
+    }
+}
+
+template <typename Node>
+Node* RB_REMOVE(RBHead<Node>* head, Node* elm) {
+    Node* child = nullptr;
+    Node* parent = nullptr;
+    Node* old = elm;
+    EntryColor color{};
+
+    const auto finalize = [&] {
+        if (color == EntryColor::Black) {
+            RB_REMOVE_COLOR(head, parent, child);
+        }
+
+        return old;
+    };
+
+    if (RB_LEFT(elm) == nullptr) {
+        child = RB_RIGHT(elm);
+    } else if (RB_RIGHT(elm) == nullptr) {
+        child = RB_LEFT(elm);
+    } else {
+        Node* left;
+        elm = RB_RIGHT(elm);
+        while ((left = RB_LEFT(elm)) != nullptr) {
+            elm = left;
+        }
+
+        child = RB_RIGHT(elm);
+        parent = RB_PARENT(elm);
+        color = RB_COLOR(elm);
+
+        if (child) {
+            RB_SET_PARENT(child, parent);
+        }
+        if (parent) {
+            if (RB_LEFT(parent) == elm) {
+                RB_SET_LEFT(parent, child);
+            } else {
+                RB_SET_RIGHT(parent, child);
+            }
+        } else {
+            head->SetRoot(child);
+        }
+
+        if (RB_PARENT(elm) == old) {
+            parent = elm;
+        }
+
+        elm->SetEntry(old->GetEntry());
+
+        if (RB_PARENT(old)) {
+            if (RB_LEFT(RB_PARENT(old)) == old) {
+                RB_SET_LEFT(RB_PARENT(old), elm);
+            } else {
+                RB_SET_RIGHT(RB_PARENT(old), elm);
+            }
+        } else {
+            head->SetRoot(elm);
+        }
+        RB_SET_PARENT(RB_LEFT(old), elm);
+        if (RB_RIGHT(old)) {
+            RB_SET_PARENT(RB_RIGHT(old), elm);
+        }
+        if (parent) {
+            left = parent;
+        }
+
+        return finalize();
+    }
+
+    parent = RB_PARENT(elm);
+    color = RB_COLOR(elm);
+
+    if (child) {
+        RB_SET_PARENT(child, parent);
+    }
+    if (parent) {
+        if (RB_LEFT(parent) == elm) {
+            RB_SET_LEFT(parent, child);
+        } else {
+            RB_SET_RIGHT(parent, child);
+        }
+    } else {
+        head->SetRoot(child);
+    }
+
+    return finalize();
+}
+
+// Inserts a node into the RB tree
+template <typename Node, typename CompareFunction>
+Node* RB_INSERT(RBHead<Node>* head, Node* elm, CompareFunction cmp) {
+    Node* parent = nullptr;
+    Node* tmp = head->Root();
     int comp = 0;
 
     while (tmp) {
@@ -550,17 +529,17 @@ constexpr T* RB_INSERT(RBHead<T>& head, T* elm, Compare cmp) {
             RB_SET_RIGHT(parent, elm);
         }
     } else {
-        head.SetRoot(elm);
+        head->SetRoot(elm);
     }
 
     RB_INSERT_COLOR(head, elm);
     return nullptr;
 }
 
-template <typename T, typename Compare>
-requires HasRBEntry<T>
-constexpr T* RB_FIND(RBHead<T>& head, T* elm, Compare cmp) {
-    T* tmp = head.Root();
+// Finds the node with the same key as elm
+template <typename Node, typename CompareFunction>
+Node* RB_FIND(RBHead<Node>* head, Node* elm, CompareFunction cmp) {
+    Node* tmp = head->Root();
 
     while (tmp) {
         const int comp = cmp(elm, tmp);
@@ -576,11 +555,11 @@ constexpr T* RB_FIND(RBHead<T>& head, T* elm, Compare cmp) {
     return nullptr;
 }
 
-template <typename T, typename Compare>
-requires HasRBEntry<T>
-constexpr T* RB_NFIND(RBHead<T>& head, T* elm, Compare cmp) {
-    T* tmp = head.Root();
-    T* res = nullptr;
+// Finds the first node greater than or equal to the search key
+template <typename Node, typename CompareFunction>
+Node* RB_NFIND(RBHead<Node>* head, Node* elm, CompareFunction cmp) {
+    Node* tmp = head->Root();
+    Node* res = nullptr;
 
     while (tmp) {
         const int comp = cmp(elm, tmp);
@@ -597,13 +576,13 @@ constexpr T* RB_NFIND(RBHead<T>& head, T* elm, Compare cmp) {
     return res;
 }
 
-template <typename T, typename U, typename Compare>
-requires HasRBEntry<T>
-constexpr T* RB_FIND_KEY(RBHead<T>& head, const U& key, Compare cmp) {
-    T* tmp = head.Root();
+// Finds the node with the same key as lelm
+template <typename Node, typename CompareFunction>
+Node* RB_FIND_LIGHT(RBHead<Node>* head, const void* lelm, CompareFunction lcmp) {
+    Node* tmp = head->Root();
 
     while (tmp) {
-        const int comp = cmp(key, tmp);
+        const int comp = lcmp(lelm, tmp);
         if (comp < 0) {
             tmp = RB_LEFT(tmp);
         } else if (comp > 0) {
@@ -616,14 +595,14 @@ constexpr T* RB_FIND_KEY(RBHead<T>& head, const U& key, Compare cmp) {
     return nullptr;
 }
 
-template <typename T, typename U, typename Compare>
-requires HasRBEntry<T>
-constexpr T* RB_NFIND_KEY(RBHead<T>& head, const U& key, Compare cmp) {
-    T* tmp = head.Root();
-    T* res = nullptr;
+// Finds the first node greater than or equal to the search key
+template <typename Node, typename CompareFunction>
+Node* RB_NFIND_LIGHT(RBHead<Node>* head, const void* lelm, CompareFunction lcmp) {
+    Node* tmp = head->Root();
+    Node* res = nullptr;
 
     while (tmp) {
-        const int comp = cmp(key, tmp);
+        const int comp = lcmp(lelm, tmp);
         if (comp < 0) {
             res = tmp;
             tmp = RB_LEFT(tmp);
@@ -637,43 +616,8 @@ constexpr T* RB_NFIND_KEY(RBHead<T>& head, const U& key, Compare cmp) {
     return res;
 }
 
-template <typename T, typename Compare>
-requires HasRBEntry<T>
-constexpr T* RB_FIND_EXISTING(RBHead<T>& head, T* elm, Compare cmp) {
-    T* tmp = head.Root();
-
-    while (true) {
-        const int comp = cmp(elm, tmp);
-        if (comp < 0) {
-            tmp = RB_LEFT(tmp);
-        } else if (comp > 0) {
-            tmp = RB_RIGHT(tmp);
-        } else {
-            return tmp;
-        }
-    }
-}
-
-template <typename T, typename U, typename Compare>
-requires HasRBEntry<T>
-constexpr T* RB_FIND_EXISTING_KEY(RBHead<T>& head, const U& key, Compare cmp) {
-    T* tmp = head.Root();
-
-    while (true) {
-        const int comp = cmp(key, tmp);
-        if (comp < 0) {
-            tmp = RB_LEFT(tmp);
-        } else if (comp > 0) {
-            tmp = RB_RIGHT(tmp);
-        } else {
-            return tmp;
-        }
-    }
-}
-
-template <typename T>
-requires HasRBEntry<T>
-constexpr T* RB_NEXT(T* elm) {
+template <typename Node>
+Node* RB_NEXT(Node* elm) {
     if (RB_RIGHT(elm)) {
         elm = RB_RIGHT(elm);
         while (RB_LEFT(elm)) {
@@ -692,9 +636,8 @@ constexpr T* RB_NEXT(T* elm) {
     return elm;
 }
 
-template <typename T>
-requires HasRBEntry<T>
-constexpr T* RB_PREV(T* elm) {
+template <typename Node>
+Node* RB_PREV(Node* elm) {
     if (RB_LEFT(elm)) {
         elm = RB_LEFT(elm);
         while (RB_RIGHT(elm)) {
@@ -713,32 +656,30 @@ constexpr T* RB_PREV(T* elm) {
     return elm;
 }
 
-template <typename T>
-requires HasRBEntry<T>
-constexpr T* RB_MIN(RBHead<T>& head) {
-    T* tmp = head.Root();
-    T* parent = nullptr;
+template <typename Node>
+Node* RB_MINMAX(RBHead<Node>* head, bool is_min) {
+    Node* tmp = head->Root();
+    Node* parent = nullptr;
 
     while (tmp) {
         parent = tmp;
-        tmp = RB_LEFT(tmp);
+        if (is_min) {
+            tmp = RB_LEFT(tmp);
+        } else {
+            tmp = RB_RIGHT(tmp);
+        }
     }
 
     return parent;
 }
 
-template <typename T>
-requires HasRBEntry<T>
-constexpr T* RB_MAX(RBHead<T>& head) {
-    T* tmp = head.Root();
-    T* parent = nullptr;
-
-    while (tmp) {
-        parent = tmp;
-        tmp = RB_RIGHT(tmp);
-    }
-
-    return parent;
+template <typename Node>
+Node* RB_MIN(RBHead<Node>* head) {
+    return RB_MINMAX(head, true);
 }
 
-} // namespace Common::freebsd
+template <typename Node>
+Node* RB_MAX(RBHead<Node>* head) {
+    return RB_MINMAX(head, false);
+}
+} // namespace Common

src/core/CMakeLists.txt

@@ -207,7 +207,6 @@ add_library(core STATIC
     hle/kernel/k_memory_region.h
     hle/kernel/k_memory_region_type.h
     hle/kernel/k_page_bitmap.h
-    hle/kernel/k_page_buffer.h
     hle/kernel/k_page_heap.cpp
     hle/kernel/k_page_heap.h
     hle/kernel/k_page_linked_list.h
@@ -245,8 +244,6 @@ add_library(core STATIC
     hle/kernel/k_system_control.h
     hle/kernel/k_thread.cpp
     hle/kernel/k_thread.h
-    hle/kernel/k_thread_local_page.cpp
-    hle/kernel/k_thread_local_page.h
     hle/kernel/k_thread_queue.cpp
     hle/kernel/k_thread_queue.h
     hle/kernel/k_trace.h

src/core/arm/dynarmic/arm_dynarmic_32.cpp

@@ -148,8 +148,8 @@ std::shared_ptr<Dynarmic::A32::Jit> ARM_Dynarmic_32::MakeJit(Common::PageTable*
     config.wall_clock_cntpct = uses_wall_clock;
 
     // Code cache size
-    config.code_cache_size = 512_MiB;
-    config.far_code_offset = 400_MiB;
+    config.code_cache_size = 128_MiB;
+    config.far_code_offset = 100_MiB;
 
     // Safe optimizations
     if (Settings::values.cpu_debug_mode) {

src/core/arm/dynarmic/arm_dynarmic_64.cpp

@@ -208,8 +208,8 @@ std::shared_ptr<Dynarmic::A64::Jit> ARM_Dynarmic_64::MakeJit(Common::PageTable*
     config.wall_clock_cntpct = uses_wall_clock;
 
     // Code cache size
-    config.code_cache_size = 512_MiB;
-    config.far_code_offset = 400_MiB;
+    config.code_cache_size = 128_MiB;
+    config.far_code_offset = 100_MiB;
 
     // Safe optimizations
     if (Settings::values.cpu_debug_mode) {

src/core/hle/ipc_helpers.h

@@ -385,7 +385,7 @@ public:
     T PopRaw();
 
     template <class T>
-    std::weak_ptr<T> PopIpcInterface() {
+    std::shared_ptr<T> PopIpcInterface() {
         ASSERT(context->Session()->IsDomain());
         ASSERT(context->GetDomainMessageHeader().input_object_count > 0);
         return context->GetDomainHandler<T>(Pop<u32>() - 1);

src/core/hle/kernel/hle_ipc.cpp

@@ -45,7 +45,7 @@ bool SessionRequestManager::HasSessionRequestHandler(const HLERequestContext& co
             LOG_CRITICAL(IPC, "object_id {} is too big!", object_id);
             return false;
         }
-        return DomainHandler(object_id - 1).lock() != nullptr;
+        return DomainHandler(object_id - 1) != nullptr;
     } else {
         return session_handler != nullptr;
     }
@@ -53,6 +53,9 @@ bool SessionRequestManager::HasSessionRequestHandler(const HLERequestContext& co
 
 void SessionRequestHandler::ClientConnected(KServerSession* session) {
     session->ClientConnected(shared_from_this());
+
+    // Ensure our server session is tracked globally.
+    kernel.RegisterServerSession(session);
 }
 
 void SessionRequestHandler::ClientDisconnected(KServerSession* session) {

src/core/hle/kernel/hle_ipc.h

@@ -94,7 +94,6 @@ protected:
     std::weak_ptr<ServiceThread> service_thread;
 };
 
-using SessionRequestHandlerWeakPtr = std::weak_ptr<SessionRequestHandler>;
 using SessionRequestHandlerPtr = std::shared_ptr<SessionRequestHandler>;
 
 /**
@@ -140,7 +139,7 @@ public:
         }
     }
 
-    SessionRequestHandlerWeakPtr DomainHandler(std::size_t index) const {
+    SessionRequestHandlerPtr DomainHandler(std::size_t index) const {
         ASSERT_MSG(index < DomainHandlerCount(), "Unexpected handler index {}", index);
         return domain_handlers.at(index);
     }
@@ -329,10 +328,10 @@ public:
 
     template <typename T>
     std::shared_ptr<T> GetDomainHandler(std::size_t index) const {
-        return std::static_pointer_cast<T>(manager.lock()->DomainHandler(index).lock());
+        return std::static_pointer_cast<T>(manager->DomainHandler(index));
     }
 
-    void SetSessionRequestManager(std::weak_ptr<SessionRequestManager> manager_) {
+    void SetSessionRequestManager(std::shared_ptr<SessionRequestManager> manager_) {
         manager = std::move(manager_);
     }
 
@@ -375,7 +374,7 @@ private:
     u32 handles_offset{};
     u32 domain_offset{};
 
-    std::weak_ptr<SessionRequestManager> manager;
+    std::shared_ptr<SessionRequestManager> manager;
 
     KernelCore& kernel;
     Core::Memory::Memory& memory;

src/core/hle/kernel/init/init_slab_setup.cpp

@@ -7,23 +7,19 @@
 #include "common/common_funcs.h"
 #include "common/common_types.h"
 #include "core/core.h"
-#include "core/device_memory.h"
 #include "core/hardware_properties.h"
 #include "core/hle/kernel/init/init_slab_setup.h"
 #include "core/hle/kernel/k_code_memory.h"
 #include "core/hle/kernel/k_event.h"
 #include "core/hle/kernel/k_memory_layout.h"
 #include "core/hle/kernel/k_memory_manager.h"
-#include "core/hle/kernel/k_page_buffer.h"
 #include "core/hle/kernel/k_port.h"
 #include "core/hle/kernel/k_process.h"
 #include "core/hle/kernel/k_resource_limit.h"
 #include "core/hle/kernel/k_session.h"
 #include "core/hle/kernel/k_shared_memory.h"
-#include "core/hle/kernel/k_shared_memory_info.h"
 #include "core/hle/kernel/k_system_control.h"
 #include "core/hle/kernel/k_thread.h"
-#include "core/hle/kernel/k_thread_local_page.h"
 #include "core/hle/kernel/k_transfer_memory.h"
 
 namespace Kernel::Init {
@@ -36,13 +32,9 @@ namespace Kernel::Init {
     HANDLER(KEvent, (SLAB_COUNT(KEvent)), ##__VA_ARGS__)                                           \
     HANDLER(KPort, (SLAB_COUNT(KPort)), ##__VA_ARGS__)                                             \
     HANDLER(KSharedMemory, (SLAB_COUNT(KSharedMemory)), ##__VA_ARGS__)                             \
-    HANDLER(KSharedMemoryInfo, (SLAB_COUNT(KSharedMemory) * 8), ##__VA_ARGS__)                     \
     HANDLER(KTransferMemory, (SLAB_COUNT(KTransferMemory)), ##__VA_ARGS__)                         \
     HANDLER(KCodeMemory, (SLAB_COUNT(KCodeMemory)), ##__VA_ARGS__)                                 \
     HANDLER(KSession, (SLAB_COUNT(KSession)), ##__VA_ARGS__)                                       \
-    HANDLER(KThreadLocalPage,                                                                      \
-            (SLAB_COUNT(KProcess) + (SLAB_COUNT(KProcess) + SLAB_COUNT(KThread)) / 8),             \
-            ##__VA_ARGS__)                                                                         \
     HANDLER(KResourceLimit, (SLAB_COUNT(KResourceLimit)), ##__VA_ARGS__)
 
 namespace {
@@ -58,46 +50,38 @@ enum KSlabType : u32 {
 // Constexpr counts.
 constexpr size_t SlabCountKProcess = 80;
 constexpr size_t SlabCountKThread = 800;
-constexpr size_t SlabCountKEvent = 900;
+constexpr size_t SlabCountKEvent = 700;
 constexpr size_t SlabCountKInterruptEvent = 100;
-constexpr size_t SlabCountKPort = 384;
+constexpr size_t SlabCountKPort = 256 + 0x20; // Extra 0x20 ports over Nintendo for homebrew.
 constexpr size_t SlabCountKSharedMemory = 80;
 constexpr size_t SlabCountKTransferMemory = 200;
 constexpr size_t SlabCountKCodeMemory = 10;
 constexpr size_t SlabCountKDeviceAddressSpace = 300;
-constexpr size_t SlabCountKSession = 1133;
+constexpr size_t SlabCountKSession = 933;
 constexpr size_t SlabCountKLightSession = 100;
 constexpr size_t SlabCountKObjectName = 7;
 constexpr size_t SlabCountKResourceLimit = 5;
 constexpr size_t SlabCountKDebug = Core::Hardware::NUM_CPU_CORES;
-constexpr size_t SlabCountKIoPool = 1;
-constexpr size_t SlabCountKIoRegion = 6;
+constexpr size_t SlabCountKAlpha = 1;
+constexpr size_t SlabCountKBeta = 6;
 
 constexpr size_t SlabCountExtraKThread = 160;
 
-/// Helper function to translate from the slab virtual address to the reserved location in physical
-/// memory.
-static PAddr TranslateSlabAddrToPhysical(KMemoryLayout& memory_layout, VAddr slab_addr) {
-    slab_addr -= memory_layout.GetSlabRegionAddress();
-    return slab_addr + Core::DramMemoryMap::SlabHeapBase;
-}
-
 template <typename T>
 VAddr InitializeSlabHeap(Core::System& system, KMemoryLayout& memory_layout, VAddr address,
                          size_t num_objects) {
+    // TODO(bunnei): This is just a place holder. We should initialize the appropriate KSlabHeap for
+    // kernel object type T with the backing kernel memory pointer once we emulate kernel memory.
 
     const size_t size = Common::AlignUp(sizeof(T) * num_objects, alignof(void*));
     VAddr start = Common::AlignUp(address, alignof(T));
 
-    // This should use the virtual memory address passed in, but currently, we do not setup the
-    // kernel virtual memory layout. Instead, we simply map these at a region of physical memory
-    // that we reserve for the slab heaps.
-    // TODO(bunnei): Fix this once we support the kernel virtual memory layout.
+    // This is intentionally empty. Once KSlabHeap is fully implemented, we can replace this with
+    // the pointer to emulated memory to pass along. Until then, KSlabHeap will just allocate/free
+    // host memory.
+    void* backing_kernel_memory{};
 
     if (size > 0) {
-        void* backing_kernel_memory{
-            system.DeviceMemory().GetPointer(TranslateSlabAddrToPhysical(memory_layout, start))};
-
         const KMemoryRegion* region = memory_layout.FindVirtual(start + size - 1);
         ASSERT(region != nullptr);
         ASSERT(region->IsDerivedFrom(KMemoryRegionType_KernelSlab));
@@ -125,8 +109,8 @@ KSlabResourceCounts KSlabResourceCounts::CreateDefault() {
         .num_KObjectName = SlabCountKObjectName,
         .num_KResourceLimit = SlabCountKResourceLimit,
         .num_KDebug = SlabCountKDebug,
-        .num_KIoPool = SlabCountKIoPool,
-        .num_KIoRegion = SlabCountKIoRegion,
+        .num_KAlpha = SlabCountKAlpha,
+        .num_KBeta = SlabCountKBeta,
     };
 }
 
@@ -137,12 +121,6 @@ void InitializeSlabResourceCounts(KernelCore& kernel) {
     }
 }
 
-size_t CalculateSlabHeapGapSize() {
-    constexpr size_t KernelSlabHeapGapSize = 2_MiB - 296_KiB;
-    static_assert(KernelSlabHeapGapSize <= KernelSlabHeapGapsSizeMax);
-    return KernelSlabHeapGapSize;
-}
-
 size_t CalculateTotalSlabHeapSize(const KernelCore& kernel) {
     size_t size = 0;
 
@@ -158,34 +136,11 @@ size_t CalculateTotalSlabHeapSize(const KernelCore& kernel) {
 #undef ADD_SLAB_SIZE
 
     // Add the reserved size.
-    size += CalculateSlabHeapGapSize();
+    size += KernelSlabHeapGapsSize;
 
     return size;
 }
 
-void InitializeKPageBufferSlabHeap(Core::System& system) {
-    auto& kernel = system.Kernel();
-
-    const auto& counts = kernel.SlabResourceCounts();
-    const size_t num_pages =
-        counts.num_KProcess + counts.num_KThread + (counts.num_KProcess + counts.num_KThread) / 8;
-    const size_t slab_size = num_pages * PageSize;
-
-    // Reserve memory from the system resource limit.
-    ASSERT(kernel.GetSystemResourceLimit()->Reserve(LimitableResource::PhysicalMemory, slab_size));
-
-    // Allocate memory for the slab.
-    constexpr auto AllocateOption = KMemoryManager::EncodeOption(
-        KMemoryManager::Pool::System, KMemoryManager::Direction::FromFront);
-    const PAddr slab_address =
-        kernel.MemoryManager().AllocateAndOpenContinuous(num_pages, 1, AllocateOption);
-    ASSERT(slab_address != 0);
-
-    // Initialize the slabheap.
-    KPageBuffer::InitializeSlabHeap(kernel, system.DeviceMemory().GetPointer(slab_address),
-                                    slab_size);
-}
-
 void InitializeSlabHeaps(Core::System& system, KMemoryLayout& memory_layout) {
     auto& kernel = system.Kernel();
 
@@ -205,9 +160,9 @@ void InitializeSlabHeaps(Core::System& system, KMemoryLayout& memory_layout) {
     }
 
     // Create an array to represent the gaps between the slabs.
-    const size_t total_gap_size = CalculateSlabHeapGapSize();
-    size_t slab_gaps[slab_types.size()];
-    for (size_t i = 0; i < slab_types.size(); i++) {
+    const size_t total_gap_size = KernelSlabHeapGapsSize;
+    std::array<size_t, slab_types.size()> slab_gaps;
+    for (size_t i = 0; i < slab_gaps.size(); i++) {
         // Note: This is an off-by-one error from Nintendo's intention, because GenerateRandomRange
         // is inclusive. However, Nintendo also has the off-by-one error, and it's "harmless", so we
         // will include it ourselves.
@@ -216,27 +171,19 @@ void InitializeSlabHeaps(Core::System& system, KMemoryLayout& memory_layout) {
 
     // Sort the array, so that we can treat differences between values as offsets to the starts of
     // slabs.
-    for (size_t i = 1; i < slab_types.size(); i++) {
+    for (size_t i = 1; i < slab_gaps.size(); i++) {
         for (size_t j = i; j > 0 && slab_gaps[j - 1] > slab_gaps[j]; j--) {
             std::swap(slab_gaps[j], slab_gaps[j - 1]);
         }
     }
 
-    // Track the gaps, so that we can free them to the unused slab tree.
-    VAddr gap_start = address;
-    size_t gap_size = 0;
-
     for (size_t i = 0; i < slab_types.size(); i++) {
         // Add the random gap to the address.
-        const auto cur_gap = (i == 0) ? slab_gaps[0] : slab_gaps[i] - slab_gaps[i - 1];
-        address += cur_gap;
-        gap_size += cur_gap;
+        address += (i == 0) ? slab_gaps[0] : slab_gaps[i] - slab_gaps[i - 1];
 
 #define INITIALIZE_SLAB_HEAP(NAME, COUNT, ...)                                                     \
     case KSlabType_##NAME:                                                                         \
-        if (COUNT > 0) {                                                                           \
-            address = InitializeSlabHeap<NAME>(system, memory_layout, address, COUNT);             \
-        }                                                                                          \
+        address = InitializeSlabHeap<NAME>(system, memory_layout, address, COUNT);                 \
         break;
 
         // Initialize the slabheap.
@@ -245,13 +192,7 @@ void InitializeSlabHeaps(Core::System& system, KMemoryLayout& memory_layout) {
             FOREACH_SLAB_TYPE(INITIALIZE_SLAB_HEAP)
             // If we somehow get an invalid type, abort.
         default:
-            UNREACHABLE_MSG("Unknown slab type: {}", slab_types[i]);
-        }
-
-        // If we've hit the end of a gap, free it.
-        if (gap_start + gap_size != address) {
-            gap_start = address;
-            gap_size = 0;
+            UNREACHABLE();
         }
     }
 }

src/core/hle/kernel/init/init_slab_setup.h

@@ -32,13 +32,12 @@ struct KSlabResourceCounts {
     size_t num_KObjectName;
     size_t num_KResourceLimit;
     size_t num_KDebug;
-    size_t num_KIoPool;
-    size_t num_KIoRegion;
+    size_t num_KAlpha;
+    size_t num_KBeta;
 };
 
 void InitializeSlabResourceCounts(KernelCore& kernel);
 size_t CalculateTotalSlabHeapSize(const KernelCore& kernel);
-void InitializeKPageBufferSlabHeap(Core::System& system);
 void InitializeSlabHeaps(Core::System& system, KMemoryLayout& memory_layout);
 
 } // namespace Kernel::Init

src/core/hle/kernel/k_address_arbiter.cpp

@@ -115,7 +115,7 @@ ResultCode KAddressArbiter::Signal(VAddr addr, s32 count) {
     {
         KScopedSchedulerLock sl(kernel);
 
-        auto it = thread_tree.nfind_key({addr, -1});
+        auto it = thread_tree.nfind_light({addr, -1});
         while ((it != thread_tree.end()) && (count <= 0 || num_waiters < count) &&
                (it->GetAddressArbiterKey() == addr)) {
             // End the thread's wait.
@@ -148,7 +148,7 @@ ResultCode KAddressArbiter::SignalAndIncrementIfEqual(VAddr addr, s32 value, s32
             return ResultInvalidState;
         }
 
-        auto it = thread_tree.nfind_key({addr, -1});
+        auto it = thread_tree.nfind_light({addr, -1});
         while ((it != thread_tree.end()) && (count <= 0 || num_waiters < count) &&
                (it->GetAddressArbiterKey() == addr)) {
             // End the thread's wait.
@@ -171,7 +171,7 @@ ResultCode KAddressArbiter::SignalAndModifyByWaitingCountIfEqual(VAddr addr, s32
     {
         [[maybe_unused]] const KScopedSchedulerLock sl(kernel);
 
-        auto it = thread_tree.nfind_key({addr, -1});
+        auto it = thread_tree.nfind_light({addr, -1});
         // Determine the updated value.
         s32 new_value{};
         if (count <= 0) {

src/core/hle/kernel/k_condition_variable.cpp

@@ -244,7 +244,7 @@ void KConditionVariable::Signal(u64 cv_key, s32 count) {
     {
         KScopedSchedulerLock sl(kernel);
 
-        auto it = thread_tree.nfind_key({cv_key, -1});
+        auto it = thread_tree.nfind_light({cv_key, -1});
         while ((it != thread_tree.end()) && (count <= 0 || num_waiters < count) &&
                (it->GetConditionVariableKey() == cv_key)) {
             KThread* target_thread = std::addressof(*it);

src/core/hle/kernel/k_memory_layout.h

@@ -57,11 +57,11 @@ constexpr std::size_t KernelPageTableHeapSize = GetMaximumOverheadSize(MainMemor
 constexpr std::size_t KernelInitialPageHeapSize = 128_KiB;
 
 constexpr std::size_t KernelSlabHeapDataSize = 5_MiB;
-constexpr std::size_t KernelSlabHeapGapsSizeMax = 2_MiB - 64_KiB;
-constexpr std::size_t KernelSlabHeapSize = KernelSlabHeapDataSize + KernelSlabHeapGapsSizeMax;
+constexpr std::size_t KernelSlabHeapGapsSize = 2_MiB - 64_KiB;
+constexpr std::size_t KernelSlabHeapSize = KernelSlabHeapDataSize + KernelSlabHeapGapsSize;
 
 // NOTE: This is calculated from KThread slab counts, assuming KThread size <= 0x860.
-constexpr std::size_t KernelSlabHeapAdditionalSize = 0x68000;
+constexpr std::size_t KernelSlabHeapAdditionalSize = 416_KiB;
 
 constexpr std::size_t KernelResourceSize =
     KernelPageTableHeapSize + KernelInitialPageHeapSize + KernelSlabHeapSize;

src/core/hle/kernel/k_page_buffer.h

@@ -1,40 +0,0 @@
-// Copyright 2022 yuzu Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#pragma once
-
-#include <array>
-
-#include "common/alignment.h"
-#include "common/assert.h"
-#include "common/common_types.h"
-#include "core/core.h"
-#include "core/device_memory.h"
-#include "core/hle/kernel/memory_types.h"
-
-namespace Kernel {
-
-class KPageBuffer final : public KSlabAllocated<KPageBuffer> {
-private:
-    alignas(PageSize) std::array<u8, PageSize> m_buffer;
-
-public:
-    KPageBuffer() {
-        std::memset(&m_buffer, 0, m_buffer.size());
-    }
-
-    PAddr GetPhysicalAddress(Core::System& system) const {
-        return system.DeviceMemory().GetPhysicalAddr(this);
-    }
-
-    static KPageBuffer* FromPhysicalAddress(Core::System& system, PAddr phys_addr) {
-        ASSERT(Common::IsAligned(phys_addr, PageSize));
-        return reinterpret_cast<KPageBuffer*>(system.DeviceMemory().GetPointer(phys_addr));
-    }
-};
-
-static_assert(sizeof(KPageBuffer) == PageSize);
-static_assert(alignof(KPageBuffer) == PageSize);
-
-} // namespace Kernel

src/core/hle/kernel/k_page_table.cpp

@@ -424,68 +424,6 @@ ResultCode KPageTable::UnmapCodeMemory(VAddr dst_address, VAddr src_address, std
     return ResultSuccess;
 }
 
-VAddr KPageTable::FindFreeArea(VAddr region_start, std::size_t region_num_pages,
-                               std::size_t num_pages, std::size_t alignment, std::size_t offset,
-                               std::size_t guard_pages) {
-    VAddr address = 0;
-
-    if (num_pages <= region_num_pages) {
-        if (this->IsAslrEnabled()) {
-            // Try to directly find a free area up to 8 times.
-            for (std::size_t i = 0; i < 8; i++) {
-                const std::size_t random_offset =
-                    KSystemControl::GenerateRandomRange(
-                        0, (region_num_pages - num_pages - guard_pages) * PageSize / alignment) *
-                    alignment;
-                const VAddr candidate =
-                    Common::AlignDown((region_start + random_offset), alignment) + offset;
-
-                KMemoryInfo info = this->QueryInfoImpl(candidate);
-
-                if (info.state != KMemoryState::Free) {
-                    continue;
-                }
-                if (!(region_start <= candidate)) {
-                    continue;
-                }
-                if (!(info.GetAddress() + guard_pages * PageSize <= candidate)) {
-                    continue;
-                }
-                if (!(candidate + (num_pages + guard_pages) * PageSize - 1 <=
-                      info.GetLastAddress())) {
-                    continue;
-                }
-                if (!(candidate + (num_pages + guard_pages) * PageSize - 1 <=
-                      region_start + region_num_pages * PageSize - 1)) {
-                    continue;
-                }
-
-                address = candidate;
-                break;
-            }
-            // Fall back to finding the first free area with a random offset.
-            if (address == 0) {
-                // NOTE: Nintendo does not account for guard pages here.
-                // This may theoretically cause an offset to be chosen that cannot be mapped. We
-                // will account for guard pages.
-                const std::size_t offset_pages = KSystemControl::GenerateRandomRange(
-                    0, region_num_pages - num_pages - guard_pages);
-                address = block_manager->FindFreeArea(region_start + offset_pages * PageSize,
-                                                      region_num_pages - offset_pages, num_pages,
-                                                      alignment, offset, guard_pages);
-            }
-        }
-
-        // Find the first free area.
-        if (address == 0) {
-            address = block_manager->FindFreeArea(region_start, region_num_pages, num_pages,
-                                                  alignment, offset, guard_pages);
-        }
-    }
-
-    return address;
-}
-
 ResultCode KPageTable::UnmapProcessMemory(VAddr dst_addr, std::size_t size,
                                           KPageTable& src_page_table, VAddr src_addr) {
     KScopedLightLock lk(general_lock);
@@ -1117,46 +1055,6 @@ ResultCode KPageTable::MapPages(VAddr address, KPageLinkedList& page_linked_list
     return ResultSuccess;
 }
 
-ResultCode KPageTable::MapPages(VAddr* out_addr, std::size_t num_pages, std::size_t alignment,
-                                PAddr phys_addr, bool is_pa_valid, VAddr region_start,
-                                std::size_t region_num_pages, KMemoryState state,
-                                KMemoryPermission perm) {
-    ASSERT(Common::IsAligned(alignment, PageSize) && alignment >= PageSize);
-
-    // Ensure this is a valid map request.
-    R_UNLESS(this->CanContain(region_start, region_num_pages * PageSize, state),
-             ResultInvalidCurrentMemory);
-    R_UNLESS(num_pages < region_num_pages, ResultOutOfMemory);
-
-    // Lock the table.
-    KScopedLightLock lk(general_lock);
-
-    // Find a random address to map at.
-    VAddr addr = this->FindFreeArea(region_start, region_num_pages, num_pages, alignment, 0,
-                                    this->GetNumGuardPages());
-    R_UNLESS(addr != 0, ResultOutOfMemory);
-    ASSERT(Common::IsAligned(addr, alignment));
-    ASSERT(this->CanContain(addr, num_pages * PageSize, state));
-    ASSERT(this->CheckMemoryState(addr, num_pages * PageSize, KMemoryState::All, KMemoryState::Free,
-                                  KMemoryPermission::None, KMemoryPermission::None,
-                                  KMemoryAttribute::None, KMemoryAttribute::None)
-               .IsSuccess());
-
-    // Perform mapping operation.
-    if (is_pa_valid) {
-        R_TRY(this->Operate(addr, num_pages, perm, OperationType::Map, phys_addr));
-    } else {
-        UNIMPLEMENTED();
-    }
-
-    // Update the blocks.
-    block_manager->Update(addr, num_pages, state, perm);
-
-    // We successfully mapped the pages.
-    *out_addr = addr;
-    return ResultSuccess;
-}
-
 ResultCode KPageTable::UnmapPages(VAddr addr, const KPageLinkedList& page_linked_list) {
     ASSERT(this->IsLockedByCurrentThread());
 
@@ -1199,30 +1097,6 @@ ResultCode KPageTable::UnmapPages(VAddr addr, KPageLinkedList& page_linked_list,
     return ResultSuccess;
 }
 
-ResultCode KPageTable::UnmapPages(VAddr address, std::size_t num_pages, KMemoryState state) {
-    // Check that the unmap is in range.
-    const std::size_t size = num_pages * PageSize;
-    R_UNLESS(this->Contains(address, size), ResultInvalidCurrentMemory);
-
-    // Lock the table.
-    KScopedLightLock lk(general_lock);
-
-    // Check the memory state.
-    std::size_t num_allocator_blocks{};
-    R_TRY(this->CheckMemoryState(std::addressof(num_allocator_blocks), address, size,
-                                 KMemoryState::All, state, KMemoryPermission::None,
-                                 KMemoryPermission::None, KMemoryAttribute::All,
-                                 KMemoryAttribute::None));
-
-    // Perform the unmap.
-    R_TRY(Operate(address, num_pages, KMemoryPermission::None, OperationType::Unmap));
-
-    // Update the blocks.
-    block_manager->Update(address, num_pages, KMemoryState::Free, KMemoryPermission::None);
-
-    return ResultSuccess;
-}
-
 ResultCode KPageTable::SetProcessMemoryPermission(VAddr addr, std::size_t size,
                                                   Svc::MemoryPermission svc_perm) {
     const size_t num_pages = size / PageSize;

src/core/hle/kernel/k_page_table.h

@@ -46,14 +46,7 @@ public:
     ResultCode UnmapMemory(VAddr dst_addr, VAddr src_addr, std::size_t size);
     ResultCode MapPages(VAddr addr, KPageLinkedList& page_linked_list, KMemoryState state,
                         KMemoryPermission perm);
-    ResultCode MapPages(VAddr* out_addr, std::size_t num_pages, std::size_t alignment,
-                        PAddr phys_addr, KMemoryState state, KMemoryPermission perm) {
-        return this->MapPages(out_addr, num_pages, alignment, phys_addr, true,
-                              this->GetRegionAddress(state), this->GetRegionSize(state) / PageSize,
-                              state, perm);
-    }
     ResultCode UnmapPages(VAddr addr, KPageLinkedList& page_linked_list, KMemoryState state);
-    ResultCode UnmapPages(VAddr address, std::size_t num_pages, KMemoryState state);
     ResultCode SetProcessMemoryPermission(VAddr addr, std::size_t size,
                                           Svc::MemoryPermission svc_perm);
     KMemoryInfo QueryInfo(VAddr addr);
@@ -98,9 +91,6 @@ private:
     ResultCode InitializeMemoryLayout(VAddr start, VAddr end);
     ResultCode MapPages(VAddr addr, const KPageLinkedList& page_linked_list,
                         KMemoryPermission perm);
-    ResultCode MapPages(VAddr* out_addr, std::size_t num_pages, std::size_t alignment,
-                        PAddr phys_addr, bool is_pa_valid, VAddr region_start,
-                        std::size_t region_num_pages, KMemoryState state, KMemoryPermission perm);
     ResultCode UnmapPages(VAddr addr, const KPageLinkedList& page_linked_list);
     bool IsRegionMapped(VAddr address, u64 size);
     bool IsRegionContiguous(VAddr addr, u64 size) const;
@@ -115,9 +105,6 @@ private:
     VAddr GetRegionAddress(KMemoryState state) const;
     std::size_t GetRegionSize(KMemoryState state) const;
 
-    VAddr FindFreeArea(VAddr region_start, std::size_t region_num_pages, std::size_t num_pages,
-                       std::size_t alignment, std::size_t offset, std::size_t guard_pages);
-
     ResultCode CheckMemoryStateContiguous(std::size_t* out_blocks_needed, VAddr addr,
                                           std::size_t size, KMemoryState state_mask,
                                           KMemoryState state, KMemoryPermission perm_mask,
@@ -150,7 +137,7 @@ private:
         return CheckMemoryState(nullptr, nullptr, nullptr, out_blocks_needed, addr, size,
                                 state_mask, state, perm_mask, perm, attr_mask, attr, ignore_attr);
     }
-    ResultCode CheckMemoryState(VAddr addr, std::size_t size, KMemoryState state_mask,
+    ResultCode CheckMemoryState(VAddr addr, size_t size, KMemoryState state_mask,
                                 KMemoryState state, KMemoryPermission perm_mask,
                                 KMemoryPermission perm, KMemoryAttribute attr_mask,
                                 KMemoryAttribute attr,
@@ -223,7 +210,7 @@ public:
     constexpr VAddr GetAliasCodeRegionSize() const {
         return alias_code_region_end - alias_code_region_start;
     }
-    std::size_t GetNormalMemorySize() {
+    size_t GetNormalMemorySize() {
         KScopedLightLock lk(general_lock);
         return GetHeapSize() + mapped_physical_memory_size;
     }

src/core/hle/kernel/k_port.cpp

@@ -57,12 +57,7 @@ ResultCode KPort::EnqueueSession(KServerSession* session) {
     R_UNLESS(state == State::Normal, ResultPortClosed);
 
     server.EnqueueSession(session);
-
-    if (auto session_ptr = server.GetSessionRequestHandler().lock(); session_ptr) {
-        session_ptr->ClientConnected(server.AcceptSession());
-    } else {
-        UNREACHABLE();
-    }
+    server.GetSessionRequestHandler()->ClientConnected(server.AcceptSession());
 
     return ResultSuccess;
 }

src/core/hle/kernel/k_process.cpp

@@ -70,6 +70,58 @@ void SetupMainThread(Core::System& system, KProcess& owner_process, u32 priority
 }
 } // Anonymous namespace
 
+// Represents a page used for thread-local storage.
+//
+// Each TLS page contains slots that may be used by processes and threads.
+// Every process and thread is created with a slot in some arbitrary page
+// (whichever page happens to have an available slot).
+class TLSPage {
+public:
+    static constexpr std::size_t num_slot_entries =
+        Core::Memory::PAGE_SIZE / Core::Memory::TLS_ENTRY_SIZE;
+
+    explicit TLSPage(VAddr address) : base_address{address} {}
+
+    bool HasAvailableSlots() const {
+        return !is_slot_used.all();
+    }
+
+    VAddr GetBaseAddress() const {
+        return base_address;
+    }
+
+    std::optional<VAddr> ReserveSlot() {
+        for (std::size_t i = 0; i < is_slot_used.size(); i++) {
+            if (is_slot_used[i]) {
+                continue;
+            }
+
+            is_slot_used[i] = true;
+            return base_address + (i * Core::Memory::TLS_ENTRY_SIZE);
+        }
+
+        return std::nullopt;
+    }
+
+    void ReleaseSlot(VAddr address) {
+        // Ensure that all given addresses are consistent with how TLS pages
+        // are intended to be used when releasing slots.
+        ASSERT(IsWithinPage(address));
+        ASSERT((address % Core::Memory::TLS_ENTRY_SIZE) == 0);
+
+        const std::size_t index = (address - base_address) / Core::Memory::TLS_ENTRY_SIZE;
+        is_slot_used[index] = false;
+    }
+
+private:
+    bool IsWithinPage(VAddr address) const {
+        return base_address <= address && address < base_address + Core::Memory::PAGE_SIZE;
+    }
+
+    VAddr base_address;
+    std::bitset<num_slot_entries> is_slot_used;
+};
+
 ResultCode KProcess::Initialize(KProcess* process, Core::System& system, std::string process_name,
                                 ProcessType type, KResourceLimit* res_limit) {
     auto& kernel = system.Kernel();
@@ -352,7 +404,7 @@ ResultCode KProcess::LoadFromMetadata(const FileSys::ProgramMetadata& metadata,
     }
 
     // Create TLS region
-    R_TRY(this->CreateThreadLocalRegion(std::addressof(tls_region_address)));
+    tls_region_address = CreateTLSRegion();
     memory_reservation.Commit();
 
     return handle_table.Initialize(capabilities.GetHandleTableSize());
@@ -392,7 +444,7 @@ void KProcess::PrepareForTermination() {
 
     stop_threads(kernel.System().GlobalSchedulerContext().GetThreadList());
 
-    this->DeleteThreadLocalRegion(tls_region_address);
+    FreeTLSRegion(tls_region_address);
     tls_region_address = 0;
 
     if (resource_limit) {
@@ -404,6 +456,9 @@ void KProcess::PrepareForTermination() {
 }
 
 void KProcess::Finalize() {
+    // Finalize the handle table and close any open handles.
+    handle_table.Finalize();
+
     // Free all shared memory infos.
     {
         auto it = shared_memory_list.begin();
@@ -428,110 +483,67 @@ void KProcess::Finalize() {
         resource_limit = nullptr;
     }
 
-    // Finalize the page table.
-    page_table.reset();
-
     // Perform inherited finalization.
     KAutoObjectWithSlabHeapAndContainer<KProcess, KWorkerTask>::Finalize();
 }
 
-ResultCode KProcess::CreateThreadLocalRegion(VAddr* out) {
-    KThreadLocalPage* tlp = nullptr;
-    VAddr tlr = 0;
-
-    // See if we can get a region from a partially used TLP.
-    {
-        KScopedSchedulerLock sl{kernel};
-
-        if (auto it = partially_used_tlp_tree.begin(); it != partially_used_tlp_tree.end()) {
-            tlr = it->Reserve();
-            ASSERT(tlr != 0);
-
-            if (it->IsAllUsed()) {
-                tlp = std::addressof(*it);
-                partially_used_tlp_tree.erase(it);
-                fully_used_tlp_tree.insert(*tlp);
-            }
-
-            *out = tlr;
-            return ResultSuccess;
-        }
-    }
-
-    // Allocate a new page.
-    tlp = KThreadLocalPage::Allocate(kernel);
-    R_UNLESS(tlp != nullptr, ResultOutOfMemory);
-    auto tlp_guard = SCOPE_GUARD({ KThreadLocalPage::Free(kernel, tlp); });
-
-    // Initialize the new page.
-    R_TRY(tlp->Initialize(kernel, this));
-
-    // Reserve a TLR.
-    tlr = tlp->Reserve();
-    ASSERT(tlr != 0);
-
-    // Insert into our tree.
-    {
-        KScopedSchedulerLock sl{kernel};
-        if (tlp->IsAllUsed()) {
-            fully_used_tlp_tree.insert(*tlp);
-        } else {
-            partially_used_tlp_tree.insert(*tlp);
-        }
-    }
-
-    // We succeeded!
-    tlp_guard.Cancel();
-    *out = tlr;
-    return ResultSuccess;
+/**
+ * Attempts to find a TLS page that contains a free slot for
+ * use by a thread.
+ *
+ * @returns If a page with an available slot is found, then an iterator
+ *          pointing to the page is returned. Otherwise the end iterator
+ *          is returned instead.
+ */
+static auto FindTLSPageWithAvailableSlots(std::vector<TLSPage>& tls_pages) {
+    return std::find_if(tls_pages.begin(), tls_pages.end(),
+                        [](const auto& page) { return page.HasAvailableSlots(); });
 }
 
-ResultCode KProcess::DeleteThreadLocalRegion(VAddr addr) {
-    KThreadLocalPage* page_to_free = nullptr;
-
-    // Release the region.
-    {
-        KScopedSchedulerLock sl{kernel};
-
-        // Try to find the page in the partially used list.
-        auto it = partially_used_tlp_tree.find_key(Common::AlignDown(addr, PageSize));
-        if (it == partially_used_tlp_tree.end()) {
-            // If we don't find it, it has to be in the fully used list.
-            it = fully_used_tlp_tree.find_key(Common::AlignDown(addr, PageSize));
-            R_UNLESS(it != fully_used_tlp_tree.end(), ResultInvalidAddress);
-
-            // Release the region.
-            it->Release(addr);
-
-            // Move the page out of the fully used list.
-            KThreadLocalPage* tlp = std::addressof(*it);
-            fully_used_tlp_tree.erase(it);
-            if (tlp->IsAllFree()) {
-                page_to_free = tlp;
-            } else {
-                partially_used_tlp_tree.insert(*tlp);
-            }
-        } else {
-            // Release the region.
-            it->Release(addr);
-
-            // Handle the all-free case.
-            KThreadLocalPage* tlp = std::addressof(*it);
-            if (tlp->IsAllFree()) {
-                partially_used_tlp_tree.erase(it);
-                page_to_free = tlp;
-            }
-        }
+VAddr KProcess::CreateTLSRegion() {
+    KScopedSchedulerLock lock(kernel);
+    if (auto tls_page_iter{FindTLSPageWithAvailableSlots(tls_pages)};
+        tls_page_iter != tls_pages.cend()) {
+        return *tls_page_iter->ReserveSlot();
     }
 
-    // If we should free the page it was in, do so.
-    if (page_to_free != nullptr) {
-        page_to_free->Finalize();
+    Page* const tls_page_ptr{kernel.GetUserSlabHeapPages().Allocate()};
+    ASSERT(tls_page_ptr);
 
-        KThreadLocalPage::Free(kernel, page_to_free);
-    }
+    const VAddr start{page_table->GetKernelMapRegionStart()};
+    const VAddr size{page_table->GetKernelMapRegionEnd() - start};
+    const PAddr tls_map_addr{kernel.System().DeviceMemory().GetPhysicalAddr(tls_page_ptr)};
+    const VAddr tls_page_addr{page_table
+                                  ->AllocateAndMapMemory(1, PageSize, true, start, size / PageSize,
+                                                         KMemoryState::ThreadLocal,
+                                                         KMemoryPermission::UserReadWrite,
+                                                         tls_map_addr)
+                                  .ValueOr(0)};
 
-    return ResultSuccess;
+    ASSERT(tls_page_addr);
+
+    std::memset(tls_page_ptr, 0, PageSize);
+    tls_pages.emplace_back(tls_page_addr);
+
+    const auto reserve_result{tls_pages.back().ReserveSlot()};
+    ASSERT(reserve_result.has_value());
+
+    return *reserve_result;
+}
+
+void KProcess::FreeTLSRegion(VAddr tls_address) {
+    KScopedSchedulerLock lock(kernel);
+    const VAddr aligned_address = Common::AlignDown(tls_address, Core::Memory::PAGE_SIZE);
+    auto iter =
+        std::find_if(tls_pages.begin(), tls_pages.end(), [aligned_address](const auto& page) {
+            return page.GetBaseAddress() == aligned_address;
+        });
+
+    // Something has gone very wrong if we're freeing a region
+    // with no actual page available.
+    ASSERT(iter != tls_pages.cend());
+
+    iter->ReleaseSlot(tls_address);
 }
 
 void KProcess::LoadModule(CodeSet code_set, VAddr base_addr) {

src/core/hle/kernel/k_process.h

@@ -15,7 +15,6 @@
 #include "core/hle/kernel/k_condition_variable.h"
 #include "core/hle/kernel/k_handle_table.h"
 #include "core/hle/kernel/k_synchronization_object.h"
-#include "core/hle/kernel/k_thread_local_page.h"
 #include "core/hle/kernel/k_worker_task.h"
 #include "core/hle/kernel/process_capability.h"
 #include "core/hle/kernel/slab_helpers.h"
@@ -363,10 +362,10 @@ public:
     // Thread-local storage management
 
     // Marks the next available region as used and returns the address of the slot.
-    [[nodiscard]] ResultCode CreateThreadLocalRegion(VAddr* out);
+    [[nodiscard]] VAddr CreateTLSRegion();
 
     // Frees a used TLS slot identified by the given address
-    ResultCode DeleteThreadLocalRegion(VAddr addr);
+    void FreeTLSRegion(VAddr tls_address);
 
 private:
     void PinThread(s32 core_id, KThread* thread) {
@@ -414,6 +413,13 @@ private:
     /// The ideal CPU core for this process, threads are scheduled on this core by default.
     u8 ideal_core = 0;
 
+    /// The Thread Local Storage area is allocated as processes create threads,
+    /// each TLS area is 0x200 bytes, so one page (0x1000) is split up in 8 parts, and each part
+    /// holds the TLS for a specific thread. This vector contains which parts are in use for each
+    /// page as a bitmask.
+    /// This vector will grow as more pages are allocated for new threads.
+    std::vector<TLSPage> tls_pages;
+
     /// Contains the parsed process capability descriptors.
     ProcessCapabilities capabilities;
 
@@ -476,12 +482,6 @@ private:
     KThread* exception_thread{};
 
     KLightLock state_lock;
-
-    using TLPTree =
-        Common::IntrusiveRedBlackTreeBaseTraits<KThreadLocalPage>::TreeType<KThreadLocalPage>;
-    using TLPIterator = TLPTree::iterator;
-    TLPTree fully_used_tlp_tree;
-    TLPTree partially_used_tlp_tree;
 };
 
 } // namespace Kernel

src/core/hle/kernel/k_server_port.h

@@ -30,11 +30,11 @@ public:
 
     /// Whether or not this server port has an HLE handler available.
     bool HasSessionRequestHandler() const {
-        return session_handler.lock() != nullptr;
+        return session_handler != nullptr;
     }
 
     /// Gets the HLE handler for this port.
-    SessionRequestHandlerWeakPtr GetSessionRequestHandler() const {
+    SessionRequestHandlerPtr GetSessionRequestHandler() const {
         return session_handler;
     }
 
@@ -42,7 +42,7 @@ public:
      * Sets the HLE handler template for the port. ServerSessions crated by connecting to this port
      * will inherit a reference to this handler.
      */
-    void SetSessionHandler(SessionRequestHandlerWeakPtr&& handler) {
+    void SetSessionHandler(SessionRequestHandlerPtr&& handler) {
         session_handler = std::move(handler);
     }
 
@@ -66,7 +66,7 @@ private:
     void CleanupSessions();
 
     SessionList session_list;
-    SessionRequestHandlerWeakPtr session_handler;
+    SessionRequestHandlerPtr session_handler;
     KPort* parent{};
 };
 

src/core/hle/kernel/k_server_session.cpp

@@ -27,7 +27,10 @@ namespace Kernel {
 
 KServerSession::KServerSession(KernelCore& kernel_) : KSynchronizationObject{kernel_} {}
 
-KServerSession::~KServerSession() = default;
+KServerSession::~KServerSession() {
+    // Ensure that the global list tracking server sessions does not hold on to a reference.
+    kernel.UnregisterServerSession(this);
+}
 
 void KServerSession::Initialize(KSession* parent_session_, std::string&& name_,
                                 std::shared_ptr<SessionRequestManager> manager_) {
@@ -46,9 +49,6 @@ void KServerSession::Destroy() {
     parent->OnServerClosed();
 
     parent->Close();
-
-    // Release host emulation members.
-    manager.reset();
 }
 
 void KServerSession::OnClientClosed() {
@@ -98,12 +98,7 @@ ResultCode KServerSession::HandleDomainSyncRequest(Kernel::HLERequestContext& co
             UNREACHABLE();
             return ResultSuccess; // Ignore error if asserts are off
         }
-        if (auto strong_ptr = manager->DomainHandler(object_id - 1).lock(); strong_ptr) {
-            return strong_ptr->HandleSyncRequest(*this, context);
-        } else {
-            UNREACHABLE();
-            return ResultSuccess;
-        }
+        return manager->DomainHandler(object_id - 1)->HandleSyncRequest(*this, context);
 
     case IPC::DomainMessageHeader::CommandType::CloseVirtualHandle: {
         LOG_DEBUG(IPC, "CloseVirtualHandle, object_id=0x{:08X}", object_id);

src/core/hle/kernel/k_slab_heap.h

@@ -16,34 +16,39 @@ class KernelCore;
 
 namespace impl {
 
-class KSlabHeapImpl {
+class KSlabHeapImpl final {
+public:
     YUZU_NON_COPYABLE(KSlabHeapImpl);
     YUZU_NON_MOVEABLE(KSlabHeapImpl);
 
-public:
     struct Node {
         Node* next{};
     };
 
-public:
     constexpr KSlabHeapImpl() = default;
+    constexpr ~KSlabHeapImpl() = default;
 
-    void Initialize() {
-        ASSERT(m_head == nullptr);
+    void Initialize(std::size_t size) {
+        ASSERT(head == nullptr);
+        obj_size = size;
+    }
+
+    constexpr std::size_t GetObjectSize() const {
+        return obj_size;
     }
 
     Node* GetHead() const {
-        return m_head;
+        return head;
     }
 
     void* Allocate() {
-        Node* ret = m_head.load();
+        Node* ret = head.load();
 
         do {
             if (ret == nullptr) {
                 break;
             }
-        } while (!m_head.compare_exchange_weak(ret, ret->next));
+        } while (!head.compare_exchange_weak(ret, ret->next));
 
         return ret;
     }
@@ -51,157 +56,170 @@ public:
     void Free(void* obj) {
         Node* node = static_cast<Node*>(obj);
 
-        Node* cur_head = m_head.load();
+        Node* cur_head = head.load();
         do {
             node->next = cur_head;
-        } while (!m_head.compare_exchange_weak(cur_head, node));
+        } while (!head.compare_exchange_weak(cur_head, node));
     }
 
 private:
-    std::atomic<Node*> m_head{};
+    std::atomic<Node*> head{};
+    std::size_t obj_size{};
 };
 
 } // namespace impl
 
-template <bool SupportDynamicExpansion>
-class KSlabHeapBase : protected impl::KSlabHeapImpl {
+class KSlabHeapBase {
+public:
     YUZU_NON_COPYABLE(KSlabHeapBase);
     YUZU_NON_MOVEABLE(KSlabHeapBase);
 
-private:
-    size_t m_obj_size{};
-    uintptr_t m_peak{};
-    uintptr_t m_start{};
-    uintptr_t m_end{};
-
-private:
-    void UpdatePeakImpl(uintptr_t obj) {
-        static_assert(std::atomic_ref<uintptr_t>::is_always_lock_free);
-        std::atomic_ref<uintptr_t> peak_ref(m_peak);
-
-        const uintptr_t alloc_peak = obj + this->GetObjectSize();
-        uintptr_t cur_peak = m_peak;
-        do {
-            if (alloc_peak <= cur_peak) {
-                break;
-            }
-        } while (!peak_ref.compare_exchange_strong(cur_peak, alloc_peak));
-    }
-
-public:
     constexpr KSlabHeapBase() = default;
+    constexpr ~KSlabHeapBase() = default;
 
-    bool Contains(uintptr_t address) const {
-        return m_start <= address && address < m_end;
+    constexpr bool Contains(uintptr_t addr) const {
+        return start <= addr && addr < end;
     }
 
-    void Initialize(size_t obj_size, void* memory, size_t memory_size) {
-        // Ensure we don't initialize a slab using null memory.
+    constexpr std::size_t GetSlabHeapSize() const {
+        return (end - start) / GetObjectSize();
+    }
+
+    constexpr std::size_t GetObjectSize() const {
+        return impl.GetObjectSize();
+    }
+
+    constexpr uintptr_t GetSlabHeapAddress() const {
+        return start;
+    }
+
+    std::size_t GetObjectIndexImpl(const void* obj) const {
+        return (reinterpret_cast<uintptr_t>(obj) - start) / GetObjectSize();
+    }
+
+    std::size_t GetPeakIndex() const {
+        return GetObjectIndexImpl(reinterpret_cast<const void*>(peak));
+    }
+
+    void* AllocateImpl() {
+        return impl.Allocate();
+    }
+
+    void FreeImpl(void* obj) {
+        // Don't allow freeing an object that wasn't allocated from this heap
+        ASSERT(Contains(reinterpret_cast<uintptr_t>(obj)));
+
+        impl.Free(obj);
+    }
+
+    void InitializeImpl(std::size_t obj_size, void* memory, std::size_t memory_size) {
+        // Ensure we don't initialize a slab using null memory
         ASSERT(memory != nullptr);
 
-        // Set our object size.
-        m_obj_size = obj_size;
+        // Initialize the base allocator
+        impl.Initialize(obj_size);
 
-        // Initialize the base allocator.
-        KSlabHeapImpl::Initialize();
+        // Set our tracking variables
+        const std::size_t num_obj = (memory_size / obj_size);
+        start = reinterpret_cast<uintptr_t>(memory);
+        end = start + num_obj * obj_size;
+        peak = start;
 
-        // Set our tracking variables.
-        const size_t num_obj = (memory_size / obj_size);
-        m_start = reinterpret_cast<uintptr_t>(memory);
-        m_end = m_start + num_obj * obj_size;
-        m_peak = m_start;
+        // Free the objects
+        u8* cur = reinterpret_cast<u8*>(end);
 
-        // Free the objects.
-        u8* cur = reinterpret_cast<u8*>(m_end);
-
-        for (size_t i = 0; i < num_obj; i++) {
+        for (std::size_t i{}; i < num_obj; i++) {
             cur -= obj_size;
-            KSlabHeapImpl::Free(cur);
+            impl.Free(cur);
         }
     }
 
-    size_t GetSlabHeapSize() const {
-        return (m_end - m_start) / this->GetObjectSize();
-    }
+private:
+    using Impl = impl::KSlabHeapImpl;
 
-    size_t GetObjectSize() const {
-        return m_obj_size;
-    }
-
-    void* Allocate() {
-        void* obj = KSlabHeapImpl::Allocate();
-
-        return obj;
-    }
-
-    void Free(void* obj) {
-        // Don't allow freeing an object that wasn't allocated from this heap.
-        const bool contained = this->Contains(reinterpret_cast<uintptr_t>(obj));
-        ASSERT(contained);
-        KSlabHeapImpl::Free(obj);
-    }
-
-    size_t GetObjectIndex(const void* obj) const {
-        if constexpr (SupportDynamicExpansion) {
-            if (!this->Contains(reinterpret_cast<uintptr_t>(obj))) {
-                return std::numeric_limits<size_t>::max();
-            }
-        }
-
-        return (reinterpret_cast<uintptr_t>(obj) - m_start) / this->GetObjectSize();
-    }
-
-    size_t GetPeakIndex() const {
-        return this->GetObjectIndex(reinterpret_cast<const void*>(m_peak));
-    }
-
-    uintptr_t GetSlabHeapAddress() const {
-        return m_start;
-    }
-
-    size_t GetNumRemaining() const {
-        // Only calculate the number of remaining objects under debug configuration.
-        return 0;
-    }
+    Impl impl;
+    uintptr_t peak{};
+    uintptr_t start{};
+    uintptr_t end{};
 };
 
 template <typename T>
-class KSlabHeap final : public KSlabHeapBase<false> {
-private:
-    using BaseHeap = KSlabHeapBase<false>;
-
+class KSlabHeap final : public KSlabHeapBase {
 public:
-    constexpr KSlabHeap() = default;
+    enum class AllocationType {
+        Host,
+        Guest,
+    };
 
-    void Initialize(void* memory, size_t memory_size) {
-        BaseHeap::Initialize(sizeof(T), memory, memory_size);
+    explicit constexpr KSlabHeap(AllocationType allocation_type_ = AllocationType::Host)
+        : KSlabHeapBase(), allocation_type{allocation_type_} {}
+
+    void Initialize(void* memory, std::size_t memory_size) {
+        if (allocation_type == AllocationType::Guest) {
+            InitializeImpl(sizeof(T), memory, memory_size);
+        }
     }
 
     T* Allocate() {
-        T* obj = static_cast<T*>(BaseHeap::Allocate());
+        switch (allocation_type) {
+        case AllocationType::Host:
+            // Fallback for cases where we do not yet support allocating guest memory from the slab
+            // heap, such as for kernel memory regions.
+            return new T;
 
-        if (obj != nullptr) [[likely]] {
-            std::construct_at(obj);
+        case AllocationType::Guest:
+            T* obj = static_cast<T*>(AllocateImpl());
+            if (obj != nullptr) {
+                new (obj) T();
+            }
+            return obj;
         }
-        return obj;
+
+        UNREACHABLE_MSG("Invalid AllocationType {}", allocation_type);
+        return nullptr;
     }
 
-    T* Allocate(KernelCore& kernel) {
-        T* obj = static_cast<T*>(BaseHeap::Allocate());
+    T* AllocateWithKernel(KernelCore& kernel) {
+        switch (allocation_type) {
+        case AllocationType::Host:
+            // Fallback for cases where we do not yet support allocating guest memory from the slab
+            // heap, such as for kernel memory regions.
+            return new T(kernel);
 
-        if (obj != nullptr) [[likely]] {
-            std::construct_at(obj, kernel);
+        case AllocationType::Guest:
+            T* obj = static_cast<T*>(AllocateImpl());
+            if (obj != nullptr) {
+                new (obj) T(kernel);
+            }
+            return obj;
         }
-        return obj;
+
+        UNREACHABLE_MSG("Invalid AllocationType {}", allocation_type);
+        return nullptr;
     }
 
     void Free(T* obj) {
-        BaseHeap::Free(obj);
+        switch (allocation_type) {
+        case AllocationType::Host:
+            // Fallback for cases where we do not yet support allocating guest memory from the slab
+            // heap, such as for kernel memory regions.
+            delete obj;
+            return;
+
+        case AllocationType::Guest:
+            FreeImpl(obj);
+            return;
+        }
+
+        UNREACHABLE_MSG("Invalid AllocationType {}", allocation_type);
     }
 
-    size_t GetObjectIndex(const T* obj) const {
-        return BaseHeap::GetObjectIndex(obj);
+    constexpr std::size_t GetObjectIndex(const T* obj) const {
+        return GetObjectIndexImpl(obj);
     }
+
+private:
+    const AllocationType allocation_type;
 };
 
 } // namespace Kernel

src/core/hle/kernel/k_thread.cpp

@@ -210,7 +210,7 @@ ResultCode KThread::Initialize(KThreadFunction func, uintptr_t arg, VAddr user_s
     if (owner != nullptr) {
         // Setup the TLS, if needed.
         if (type == ThreadType::User) {
-            R_TRY(owner->CreateThreadLocalRegion(std::addressof(tls_address)));
+            tls_address = owner->CreateTLSRegion();
         }
 
         parent = owner;
@@ -305,7 +305,7 @@ void KThread::Finalize() {
 
     // If the thread has a local region, delete it.
     if (tls_address != 0) {
-        ASSERT(parent->DeleteThreadLocalRegion(tls_address).IsSuccess());
+        parent->FreeTLSRegion(tls_address);
     }
 
     // Release any waiters.
@@ -326,9 +326,6 @@ void KThread::Finalize() {
         }
     }
 
-    // Release host emulation members.
-    host_context.reset();
-
     // Perform inherited finalization.
     KSynchronizationObject::Finalize();
 }

src/core/hle/kernel/k_thread.h

@@ -656,7 +656,7 @@ private:
     static_assert(sizeof(SyncObjectBuffer::sync_objects) == sizeof(SyncObjectBuffer::handles));
 
     struct ConditionVariableComparator {
-        struct RedBlackKeyType {
+        struct LightCompareType {
             u64 cv_key{};
             s32 priority{};
 
@@ -672,8 +672,8 @@ private:
         template <typename T>
         requires(
             std::same_as<T, KThread> ||
-            std::same_as<T, RedBlackKeyType>) static constexpr int Compare(const T& lhs,
-                                                                           const KThread& rhs) {
+            std::same_as<T, LightCompareType>) static constexpr int Compare(const T& lhs,
+                                                                            const KThread& rhs) {
             const u64 l_key = lhs.GetConditionVariableKey();
             const u64 r_key = rhs.GetConditionVariableKey();
 

src/core/hle/kernel/k_thread_local_page.cpp

@@ -1,65 +0,0 @@
-// Copyright 2022 yuzu Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#include "common/scope_exit.h"
-#include "core/hle/kernel/k_memory_block.h"
-#include "core/hle/kernel/k_page_table.h"
-#include "core/hle/kernel/k_process.h"
-#include "core/hle/kernel/k_thread_local_page.h"
-#include "core/hle/kernel/kernel.h"
-
-namespace Kernel {
-
-ResultCode KThreadLocalPage::Initialize(KernelCore& kernel, KProcess* process) {
-    // Set that this process owns us.
-    m_owner = process;
-    m_kernel = &kernel;
-
-    // Allocate a new page.
-    KPageBuffer* page_buf = KPageBuffer::Allocate(kernel);
-    R_UNLESS(page_buf != nullptr, ResultOutOfMemory);
-    auto page_buf_guard = SCOPE_GUARD({ KPageBuffer::Free(kernel, page_buf); });
-
-    // Map the address in.
-    const auto phys_addr = kernel.System().DeviceMemory().GetPhysicalAddr(page_buf);
-    R_TRY(m_owner->PageTable().MapPages(std::addressof(m_virt_addr), 1, PageSize, phys_addr,
-                                        KMemoryState::ThreadLocal,
-                                        KMemoryPermission::UserReadWrite));
-
-    // We succeeded.
-    page_buf_guard.Cancel();
-
-    return ResultSuccess;
-}
-
-ResultCode KThreadLocalPage::Finalize() {
-    // Get the physical address of the page.
-    const PAddr phys_addr = m_owner->PageTable().GetPhysicalAddr(m_virt_addr);
-    ASSERT(phys_addr);
-
-    // Unmap the page.
-    R_TRY(m_owner->PageTable().UnmapPages(this->GetAddress(), 1, KMemoryState::ThreadLocal));
-
-    // Free the page.
-    KPageBuffer::Free(*m_kernel, KPageBuffer::FromPhysicalAddress(m_kernel->System(), phys_addr));
-
-    return ResultSuccess;
-}
-
-VAddr KThreadLocalPage::Reserve() {
-    for (size_t i = 0; i < m_is_region_free.size(); i++) {
-        if (m_is_region_free[i]) {
-            m_is_region_free[i] = false;
-            return this->GetRegionAddress(i);
-        }
-    }
-
-    return 0;
-}
-
-void KThreadLocalPage::Release(VAddr addr) {
-    m_is_region_free[this->GetRegionIndex(addr)] = true;
-}
-
-} // namespace Kernel

src/core/hle/kernel/k_thread_local_page.h

@@ -1,121 +0,0 @@
-// Copyright 2022 yuzu Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#pragma once
-
-#include <array>
-
-#include "common/alignment.h"
-#include "common/assert.h"
-#include "common/common_types.h"
-#include "common/intrusive_red_black_tree.h"
-#include "core/hle/kernel/k_page_buffer.h"
-#include "core/hle/kernel/memory_types.h"
-#include "core/hle/kernel/slab_helpers.h"
-#include "core/hle/result.h"
-
-namespace Kernel {
-
-class KernelCore;
-class KProcess;
-
-class KThreadLocalPage final : public Common::IntrusiveRedBlackTreeBaseNode<KThreadLocalPage>,
-                               public KSlabAllocated<KThreadLocalPage> {
-public:
-    static constexpr size_t RegionsPerPage = PageSize / Svc::ThreadLocalRegionSize;
-    static_assert(RegionsPerPage > 0);
-
-public:
-    explicit KThreadLocalPage(VAddr addr = {}) : m_virt_addr(addr) {
-        for (size_t i = 0; i < m_is_region_free.size(); i++) {
-            m_is_region_free[i] = true;
-        }
-    }
-
-    constexpr VAddr GetAddress() const {
-        return m_virt_addr;
-    }
-
-    ResultCode Initialize(KernelCore& kernel, KProcess* process);
-    ResultCode Finalize();
-
-    VAddr Reserve();
-    void Release(VAddr addr);
-
-    bool IsAllUsed() const {
-        for (size_t i = 0; i < RegionsPerPage; i++) {
-            if (m_is_region_free[i]) {
-                return false;
-            }
-        }
-        return true;
-    }
-
-    bool IsAllFree() const {
-        for (size_t i = 0; i < RegionsPerPage; i++) {
-            if (!m_is_region_free[i]) {
-                return false;
-            }
-        }
-        return true;
-    }
-
-    bool IsAnyUsed() const {
-        return !this->IsAllFree();
-    }
-
-    bool IsAnyFree() const {
-        return !this->IsAllUsed();
-    }
-
-public:
-    using RedBlackKeyType = VAddr;
-
-    static constexpr RedBlackKeyType GetRedBlackKey(const RedBlackKeyType& v) {
-        return v;
-    }
-    static constexpr RedBlackKeyType GetRedBlackKey(const KThreadLocalPage& v) {
-        return v.GetAddress();
-    }
-
-    template <typename T>
-    requires(std::same_as<T, KThreadLocalPage> ||
-             std::same_as<T, RedBlackKeyType>) static constexpr int Compare(const T& lhs,
-                                                                            const KThreadLocalPage&
-                                                                                rhs) {
-        const VAddr lval = GetRedBlackKey(lhs);
-        const VAddr rval = GetRedBlackKey(rhs);
-
-        if (lval < rval) {
-            return -1;
-        } else if (lval == rval) {
-            return 0;
-        } else {
-            return 1;
-        }
-    }
-
-private:
-    constexpr VAddr GetRegionAddress(size_t i) {
-        return this->GetAddress() + i * Svc::ThreadLocalRegionSize;
-    }
-
-    constexpr bool Contains(VAddr addr) {
-        return this->GetAddress() <= addr && addr < this->GetAddress() + PageSize;
-    }
-
-    constexpr size_t GetRegionIndex(VAddr addr) {
-        ASSERT(Common::IsAligned(addr, Svc::ThreadLocalRegionSize));
-        ASSERT(this->Contains(addr));
-        return (addr - this->GetAddress()) / Svc::ThreadLocalRegionSize;
-    }
-
-private:
-    VAddr m_virt_addr{};
-    KProcess* m_owner{};
-    KernelCore* m_kernel{};
-    std::array<bool, RegionsPerPage> m_is_region_free{};
-};
-
-} // namespace Kernel

src/core/hle/kernel/kernel.cpp

@@ -52,7 +52,7 @@ namespace Kernel {
 
 struct KernelCore::Impl {
     explicit Impl(Core::System& system_, KernelCore& kernel_)
-        : time_manager{system_},
+        : time_manager{system_}, object_list_container{kernel_},
           service_threads_manager{1, "yuzu:ServiceThreadsManager"}, system{system_} {}
 
     void SetMulticore(bool is_multi) {
@@ -60,7 +60,6 @@ struct KernelCore::Impl {
     }
 
     void Initialize(KernelCore& kernel) {
-        global_object_list_container = std::make_unique<KAutoObjectWithListContainer>(kernel);
         global_scheduler_context = std::make_unique<Kernel::GlobalSchedulerContext>(kernel);
         global_handle_table = std::make_unique<Kernel::KHandleTable>(kernel);
         global_handle_table->Initialize(KHandleTable::MaxTableSize);
@@ -77,7 +76,7 @@ struct KernelCore::Impl {
         // Initialize kernel memory and resources.
         InitializeSystemResourceLimit(kernel, system.CoreTiming());
         InitializeMemoryLayout();
-        Init::InitializeKPageBufferSlabHeap(system);
+        InitializePageSlab();
         InitializeSchedulers();
         InitializeSuspendThreads();
         InitializePreemption(kernel);
@@ -108,6 +107,19 @@ struct KernelCore::Impl {
         for (auto* server_port : server_ports_) {
             server_port->Close();
         }
+        // Close all open server sessions.
+        std::unordered_set<KServerSession*> server_sessions_;
+        {
+            std::lock_guard lk(server_sessions_lock);
+            server_sessions_ = server_sessions;
+            server_sessions.clear();
+        }
+        for (auto* server_session : server_sessions_) {
+            server_session->Close();
+        }
+
+        // Ensure that the object list container is finalized and properly shutdown.
+        object_list_container.Finalize();
 
         // Ensures all service threads gracefully shutdown.
         ClearServiceThreads();
@@ -182,15 +194,11 @@ struct KernelCore::Impl {
         {
             std::lock_guard lk(registered_objects_lock);
             if (registered_objects.size()) {
-                LOG_DEBUG(Kernel, "{} kernel objects were dangling on shutdown!",
-                          registered_objects.size());
+                LOG_WARNING(Kernel, "{} kernel objects were dangling on shutdown!",
+                            registered_objects.size());
                 registered_objects.clear();
             }
         }
-
-        // Ensure that the object list container is finalized and properly shutdown.
-        global_object_list_container->Finalize();
-        global_object_list_container.reset();
     }
 
     void InitializePhysicalCores() {
@@ -283,16 +291,15 @@ struct KernelCore::Impl {
 
     // Gets the dummy KThread for the caller, allocating a new one if this is the first time
     KThread* GetHostDummyThread() {
-        auto initialize = [this](KThread* thread) {
+        auto make_thread = [this]() {
+            KThread* thread = KThread::Create(system.Kernel());
             ASSERT(KThread::InitializeDummyThread(thread).IsSuccess());
             thread->SetName(fmt::format("DummyThread:{}", GetHostThreadId()));
             return thread;
         };
 
-        thread_local auto raw_thread = KThread(system.Kernel());
-        thread_local auto thread = initialize(&raw_thread);
-
-        return thread;
+        thread_local KThread* saved_thread = make_thread();
+        return saved_thread;
     }
 
     /// Registers a CPU core thread by allocating a host thread ID for it
@@ -653,6 +660,22 @@ struct KernelCore::Impl {
                                     time_phys_addr, time_size, "Time:SharedMemory");
     }
 
+    void InitializePageSlab() {
+        // Allocate slab heaps
+        user_slab_heap_pages =
+            std::make_unique<KSlabHeap<Page>>(KSlabHeap<Page>::AllocationType::Guest);
+
+        // TODO(ameerj): This should be derived, not hardcoded within the kernel
+        constexpr u64 user_slab_heap_size{0x3de000};
+        // Reserve slab heaps
+        ASSERT(
+            system_resource_limit->Reserve(LimitableResource::PhysicalMemory, user_slab_heap_size));
+        // Initialize slab heap
+        user_slab_heap_pages->Initialize(
+            system.DeviceMemory().GetPointer(Core::DramMemoryMap::SlabHeapBase),
+            user_slab_heap_size);
+    }
+
     KClientPort* CreateNamedServicePort(std::string name) {
         auto search = service_interface_factory.find(name);
         if (search == service_interface_factory.end()) {
@@ -690,6 +713,7 @@ struct KernelCore::Impl {
     }
 
     std::mutex server_ports_lock;
+    std::mutex server_sessions_lock;
     std::mutex registered_objects_lock;
     std::mutex registered_in_use_objects_lock;
 
@@ -713,13 +737,14 @@ struct KernelCore::Impl {
     // stores all the objects in place.
     std::unique_ptr<KHandleTable> global_handle_table;
 
-    std::unique_ptr<KAutoObjectWithListContainer> global_object_list_container;
+    KAutoObjectWithListContainer object_list_container;
 
     /// Map of named ports managed by the kernel, which can be retrieved using
     /// the ConnectToPort SVC.
     std::unordered_map<std::string, ServiceInterfaceFactory> service_interface_factory;
     NamedPortTable named_ports;
     std::unordered_set<KServerPort*> server_ports;
+    std::unordered_set<KServerSession*> server_sessions;
     std::unordered_set<KAutoObject*> registered_objects;
     std::unordered_set<KAutoObject*> registered_in_use_objects;
 
@@ -731,6 +756,7 @@ struct KernelCore::Impl {
 
     // Kernel memory management
     std::unique_ptr<KMemoryManager> memory_manager;
+    std::unique_ptr<KSlabHeap<Page>> user_slab_heap_pages;
 
     // Shared memory for services
     Kernel::KSharedMemory* hid_shared_mem{};
@@ -889,11 +915,11 @@ const Core::ExclusiveMonitor& KernelCore::GetExclusiveMonitor() const {
 }
 
 KAutoObjectWithListContainer& KernelCore::ObjectListContainer() {
-    return *impl->global_object_list_container;
+    return impl->object_list_container;
 }
 
 const KAutoObjectWithListContainer& KernelCore::ObjectListContainer() const {
-    return *impl->global_object_list_container;
+    return impl->object_list_container;
 }
 
 void KernelCore::InvalidateAllInstructionCaches() {
@@ -923,6 +949,16 @@ KClientPort* KernelCore::CreateNamedServicePort(std::string name) {
     return impl->CreateNamedServicePort(std::move(name));
 }
 
+void KernelCore::RegisterServerSession(KServerSession* server_session) {
+    std::lock_guard lk(impl->server_sessions_lock);
+    impl->server_sessions.insert(server_session);
+}
+
+void KernelCore::UnregisterServerSession(KServerSession* server_session) {
+    std::lock_guard lk(impl->server_sessions_lock);
+    impl->server_sessions.erase(server_session);
+}
+
 void KernelCore::RegisterKernelObject(KAutoObject* object) {
     std::lock_guard lk(impl->registered_objects_lock);
     impl->registered_objects.insert(object);
@@ -995,6 +1031,14 @@ const KMemoryManager& KernelCore::MemoryManager() const {
     return *impl->memory_manager;
 }
 
+KSlabHeap<Page>& KernelCore::GetUserSlabHeapPages() {
+    return *impl->user_slab_heap_pages;
+}
+
+const KSlabHeap<Page>& KernelCore::GetUserSlabHeapPages() const {
+    return *impl->user_slab_heap_pages;
+}
+
 Kernel::KSharedMemory& KernelCore::GetHidSharedMem() {
     return *impl->hid_shared_mem;
 }

src/core/hle/kernel/kernel.h

@@ -43,7 +43,6 @@ class KHandleTable;
 class KLinkedListNode;
 class KMemoryLayout;
 class KMemoryManager;
-class KPageBuffer;
 class KPort;
 class KProcess;
 class KResourceLimit;
@@ -53,7 +52,6 @@ class KSession;
 class KSharedMemory;
 class KSharedMemoryInfo;
 class KThread;
-class KThreadLocalPage;
 class KTransferMemory;
 class KWorkerTaskManager;
 class KWritableEvent;
@@ -196,6 +194,14 @@ public:
     /// Opens a port to a service previously registered with RegisterNamedService.
     KClientPort* CreateNamedServicePort(std::string name);
 
+    /// Registers a server session with the gobal emulation state, to be freed on shutdown. This is
+    /// necessary because we do not emulate processes for HLE sessions.
+    void RegisterServerSession(KServerSession* server_session);
+
+    /// Unregisters a server session previously registered with RegisterServerSession when it was
+    /// destroyed during the current emulation session.
+    void UnregisterServerSession(KServerSession* server_session);
+
     /// Registers all kernel objects with the global emulation state, this is purely for tracking
     /// leaks after emulation has been shutdown.
     void RegisterKernelObject(KAutoObject* object);
@@ -233,6 +239,12 @@ public:
     /// Gets the virtual memory manager for the kernel.
     const KMemoryManager& MemoryManager() const;
 
+    /// Gets the slab heap allocated for user space pages.
+    KSlabHeap<Page>& GetUserSlabHeapPages();
+
+    /// Gets the slab heap allocated for user space pages.
+    const KSlabHeap<Page>& GetUserSlabHeapPages() const;
+
     /// Gets the shared memory object for HID services.
     Kernel::KSharedMemory& GetHidSharedMem();
 
@@ -324,10 +336,6 @@ public:
             return slab_heap_container->writeable_event;
         } else if constexpr (std::is_same_v<T, KCodeMemory>) {
             return slab_heap_container->code_memory;
-        } else if constexpr (std::is_same_v<T, KPageBuffer>) {
-            return slab_heap_container->page_buffer;
-        } else if constexpr (std::is_same_v<T, KThreadLocalPage>) {
-            return slab_heap_container->thread_local_page;
         }
     }
 
@@ -389,8 +397,6 @@ private:
         KSlabHeap<KTransferMemory> transfer_memory;
         KSlabHeap<KWritableEvent> writeable_event;
         KSlabHeap<KCodeMemory> code_memory;
-        KSlabHeap<KPageBuffer> page_buffer;
-        KSlabHeap<KThreadLocalPage> thread_local_page;
     };
 
     std::unique_ptr<SlabHeapContainer> slab_heap_container;

src/core/hle/kernel/service_thread.cpp

@@ -49,9 +49,12 @@ ServiceThread::Impl::Impl(KernelCore& kernel, std::size_t num_threads, const std
                 return;
             }
 
-            // Allocate a dummy guest thread for this host thread.
             kernel.RegisterHostThread();
 
+            // Ensure the dummy thread allocated for this host thread is closed on exit.
+            auto* dummy_thread = kernel.GetCurrentEmuThread();
+            SCOPE_EXIT({ dummy_thread->Close(); });
+
             while (true) {
                 std::function<void()> task;
 

src/core/hle/kernel/slab_helpers.h

@@ -59,7 +59,7 @@ class KAutoObjectWithSlabHeapAndContainer : public Base {
 
 private:
     static Derived* Allocate(KernelCore& kernel) {
-        return kernel.SlabHeap<Derived>().Allocate(kernel);
+        return kernel.SlabHeap<Derived>().AllocateWithKernel(kernel);
     }
 
     static void Free(KernelCore& kernel, Derived* obj) {

src/core/hle/kernel/svc_types.h

@@ -96,6 +96,4 @@ constexpr inline s32 IdealCoreNoUpdate = -3;
 constexpr inline s32 LowestThreadPriority = 63;
 constexpr inline s32 HighestThreadPriority = 0;
 
-constexpr inline size_t ThreadLocalRegionSize = 0x200;
-
 } // namespace Kernel::Svc

src/core/hle/service/am/am.cpp

@@ -980,7 +980,7 @@ private:
         LOG_DEBUG(Service_AM, "called");
 
         IPC::RequestParser rp{ctx};
-        applet->GetBroker().PushNormalDataFromGame(rp.PopIpcInterface<IStorage>().lock());
+        applet->GetBroker().PushNormalDataFromGame(rp.PopIpcInterface<IStorage>());
 
         IPC::ResponseBuilder rb{ctx, 2};
         rb.Push(ResultSuccess);
@@ -1007,7 +1007,7 @@ private:
         LOG_DEBUG(Service_AM, "called");
 
         IPC::RequestParser rp{ctx};
-        applet->GetBroker().PushInteractiveDataFromGame(rp.PopIpcInterface<IStorage>().lock());
+        applet->GetBroker().PushInteractiveDataFromGame(rp.PopIpcInterface<IStorage>());
 
         ASSERT(applet->IsInitialized());
         applet->ExecuteInteractive();

src/core/hle/service/kernel_helpers.cpp

@@ -17,12 +17,21 @@ namespace Service::KernelHelpers {
 
 ServiceContext::ServiceContext(Core::System& system_, std::string name_)
     : kernel(system_.Kernel()) {
+
+    // Create a resource limit for the process.
+    const auto physical_memory_size =
+        kernel.MemoryManager().GetSize(Kernel::KMemoryManager::Pool::System);
+    auto* resource_limit = Kernel::CreateResourceLimitForProcess(system_, physical_memory_size);
+
     // Create the process.
     process = Kernel::KProcess::Create(kernel);
     ASSERT(Kernel::KProcess::Initialize(process, system_, std::move(name_),
                                         Kernel::KProcess::ProcessType::KernelInternal,
-                                        kernel.GetSystemResourceLimit())
+                                        resource_limit)
                .IsSuccess());
+
+    // Close reference to our resource limit, as the process opens one.
+    resource_limit->Close();
 }
 
 ServiceContext::~ServiceContext() {

src/core/hle/service/sm/sm.cpp

@@ -81,8 +81,6 @@ ResultVal<Kernel::KPort*> ServiceManager::GetServicePort(const std::string& name
     }
 
     auto* port = Kernel::KPort::Create(kernel);
-    SCOPE_EXIT({ port->Close(); });
-
     port->Initialize(ServerSessionCountMax, false, name);
     auto handler = it->second;
     port->GetServerPort().SetSessionHandler(std::move(handler));

src/shader_recompiler/backend/spirv/emit_spirv.h

@@ -22,7 +22,7 @@ constexpr u32 NUM_TEXTURE_AND_IMAGE_SCALING_WORDS =
 struct RescalingLayout {
     alignas(16) std::array<u32, NUM_TEXTURE_SCALING_WORDS> rescaling_textures;
     alignas(16) std::array<u32, NUM_IMAGE_SCALING_WORDS> rescaling_images;
-    alignas(16) u32 down_factor;
+    u32 down_factor;
 };
 constexpr u32 RESCALING_LAYOUT_WORDS_OFFSET = offsetof(RescalingLayout, rescaling_textures);
 constexpr u32 RESCALING_LAYOUT_DOWN_FACTOR_OFFSET = offsetof(RescalingLayout, down_factor);

src/shader_recompiler/frontend/maxwell/translate_program.cpp

@@ -212,11 +212,11 @@ IR::Program TranslateProgram(ObjectPool<IR::Inst>& inst_pool, ObjectPool<IR::Blo
     }
     Optimization::SsaRewritePass(program);
 
+    Optimization::ConstantPropagationPass(program);
+
     Optimization::GlobalMemoryToStorageBufferPass(program);
     Optimization::TexturePass(env, program);
 
-    Optimization::ConstantPropagationPass(program);
-
     if (Settings::values.resolution_info.active) {
         Optimization::RescalingPass(program);
     }

src/shader_recompiler/ir_opt/global_memory_to_storage_buffer_pass.cpp

@@ -334,7 +334,8 @@ std::optional<LowAddrInfo> TrackLowAddress(IR::Inst* inst) {
 /// Tries to track the storage buffer address used by a global memory instruction
 std::optional<StorageBufferAddr> Track(const IR::Value& value, const Bias* bias) {
     const auto pred{[bias](const IR::Inst* inst) -> std::optional<StorageBufferAddr> {
-        if (inst->GetOpcode() != IR::Opcode::GetCbufU32) {
+        if (inst->GetOpcode() != IR::Opcode::GetCbufU32 &&
+            inst->GetOpcode() != IR::Opcode::GetCbufU32x2) {
             return std::nullopt;
         }
         const IR::Value index{inst->Arg(0)};

src/shader_recompiler/ir_opt/rescaling_pass.cpp

@@ -183,6 +183,31 @@ void ScaleIntegerComposite(IR::IREmitter& ir, IR::Inst& inst, const IR::U1& is_s
     }
 }
 
+void ScaleIntegerOffsetComposite(IR::IREmitter& ir, IR::Inst& inst, const IR::U1& is_scaled,
+                                 size_t index) {
+    const IR::Value composite{inst.Arg(index)};
+    if (composite.IsEmpty()) {
+        return;
+    }
+    const auto info{inst.Flags<IR::TextureInstInfo>()};
+    const IR::U32 x{Scale(ir, is_scaled, IR::U32{ir.CompositeExtract(composite, 0)})};
+    const IR::U32 y{Scale(ir, is_scaled, IR::U32{ir.CompositeExtract(composite, 1)})};
+    switch (info.type) {
+    case TextureType::ColorArray2D:
+    case TextureType::Color2D:
+        inst.SetArg(index, ir.CompositeConstruct(x, y));
+        break;
+    case TextureType::Color1D:
+    case TextureType::ColorArray1D:
+    case TextureType::Color3D:
+    case TextureType::ColorCube:
+    case TextureType::ColorArrayCube:
+    case TextureType::Buffer:
+        // Nothing to patch here
+        break;
+    }
+}
+
 void SubScaleCoord(IR::IREmitter& ir, IR::Inst& inst, const IR::U1& is_scaled) {
     const auto info{inst.Flags<IR::TextureInstInfo>()};
     const IR::Value coord{inst.Arg(1)};
@@ -220,7 +245,7 @@ void SubScaleImageFetch(IR::Block& block, IR::Inst& inst) {
     const IR::U1 is_scaled{ir.IsTextureScaled(ir.Imm32(info.descriptor_index))};
     SubScaleCoord(ir, inst, is_scaled);
     // Scale ImageFetch offset
-    ScaleIntegerComposite(ir, inst, is_scaled, 2);
+    ScaleIntegerOffsetComposite(ir, inst, is_scaled, 2);
 }
 
 void SubScaleImageRead(IR::Block& block, IR::Inst& inst) {
@@ -242,7 +267,7 @@ void PatchImageFetch(IR::Block& block, IR::Inst& inst) {
     const IR::U1 is_scaled{ir.IsTextureScaled(ir.Imm32(info.descriptor_index))};
     ScaleIntegerComposite(ir, inst, is_scaled, 1);
     // Scale ImageFetch offset
-    ScaleIntegerComposite(ir, inst, is_scaled, 2);
+    ScaleIntegerOffsetComposite(ir, inst, is_scaled, 2);
 }
 
 void PatchImageRead(IR::Block& block, IR::Inst& inst) {

src/video_core/engines/maxwell_3d.cpp

@@ -7,6 +7,7 @@
 #include "common/assert.h"
 #include "core/core.h"
 #include "core/core_timing.h"
+#include "video_core/dirty_flags.h"
 #include "video_core/engines/maxwell_3d.h"
 #include "video_core/gpu.h"
 #include "video_core/memory_manager.h"
@@ -208,6 +209,14 @@ void Maxwell3D::ProcessMethodCall(u32 method, u32 argument, u32 nonshadow_argume
         return ProcessCBBind(4);
     case MAXWELL3D_REG_INDEX(draw.vertex_end_gl):
         return DrawArrays();
+    case MAXWELL3D_REG_INDEX(small_index):
+        regs.index_array.count = regs.small_index.count;
+        regs.index_array.first = regs.small_index.first;
+        dirty.flags[VideoCommon::Dirty::IndexBuffer] = true;
+        return DrawArrays();
+    case MAXWELL3D_REG_INDEX(topology_override):
+        use_topology_override = true;
+        return;
     case MAXWELL3D_REG_INDEX(clear_buffers):
         return ProcessClearBuffers();
     case MAXWELL3D_REG_INDEX(query.query_get):
@@ -360,6 +369,35 @@ void Maxwell3D::CallMethodFromMME(u32 method, u32 method_argument) {
     }
 }
 
+void Maxwell3D::ProcessTopologyOverride() {
+    using PrimitiveTopology = Maxwell3D::Regs::PrimitiveTopology;
+    using PrimitiveTopologyOverride = Maxwell3D::Regs::PrimitiveTopologyOverride;
+
+    PrimitiveTopology topology{};
+
+    switch (regs.topology_override) {
+    case PrimitiveTopologyOverride::None:
+        topology = regs.draw.topology;
+        break;
+    case PrimitiveTopologyOverride::Points:
+        topology = PrimitiveTopology::Points;
+        break;
+    case PrimitiveTopologyOverride::Lines:
+        topology = PrimitiveTopology::Lines;
+        break;
+    case PrimitiveTopologyOverride::LineStrip:
+        topology = PrimitiveTopology::LineStrip;
+        break;
+    default:
+        topology = static_cast<PrimitiveTopology>(regs.topology_override);
+        break;
+    }
+
+    if (use_topology_override) {
+        regs.draw.topology.Assign(topology);
+    }
+}
+
 void Maxwell3D::FlushMMEInlineDraw() {
     LOG_TRACE(HW_GPU, "called, topology={}, count={}", regs.draw.topology.Value(),
               regs.vertex_buffer.count);
@@ -370,6 +408,8 @@ void Maxwell3D::FlushMMEInlineDraw() {
     ASSERT_MSG(!regs.draw.instance_next || !regs.draw.instance_cont,
                "Illegal combination of instancing parameters");
 
+    ProcessTopologyOverride();
+
     const bool is_indexed = mme_draw.current_mode == MMEDrawMode::Indexed;
     if (ShouldExecute()) {
         rasterizer->Draw(is_indexed, true);
@@ -529,6 +569,8 @@ void Maxwell3D::DrawArrays() {
     ASSERT_MSG(!regs.draw.instance_next || !regs.draw.instance_cont,
                "Illegal combination of instancing parameters");
 
+    ProcessTopologyOverride();
+
     if (regs.draw.instance_next) {
         // Increment the current instance *before* drawing.
         state.current_instance += 1;

src/video_core/engines/maxwell_3d.h

@@ -367,6 +367,22 @@ public:
             Patches = 0xe,
         };
 
+        // Constants as from NVC0_3D_UNK1970_D3D
+        // https://gitlab.freedesktop.org/mesa/mesa/-/blob/main/src/gallium/drivers/nouveau/nvc0/nvc0_3d.xml.h#L1598
+        enum class PrimitiveTopologyOverride : u32 {
+            None = 0x0,
+            Points = 0x1,
+            Lines = 0x2,
+            LineStrip = 0x3,
+            Triangles = 0x4,
+            TriangleStrip = 0x5,
+            LinesAdjacency = 0xa,
+            LineStripAdjacency = 0xb,
+            TrianglesAdjacency = 0xc,
+            TriangleStripAdjacency = 0xd,
+            Patches = 0xe,
+        };
+
         enum class IndexFormat : u32 {
             UnsignedByte = 0x0,
             UnsignedShort = 0x1,
@@ -1200,7 +1216,12 @@ public:
                     }
                 } index_array;
 
-                INSERT_PADDING_WORDS_NOINIT(0x7);
+                union {
+                    BitField<0, 16, u32> first;
+                    BitField<16, 16, u32> count;
+                } small_index;
+
+                INSERT_PADDING_WORDS_NOINIT(0x6);
 
                 INSERT_PADDING_WORDS_NOINIT(0x1F);
 
@@ -1244,7 +1265,11 @@ public:
                     BitField<11, 1, u32> depth_clamp_disabled;
                 } view_volume_clip_control;
 
-                INSERT_PADDING_WORDS_NOINIT(0x1F);
+                INSERT_PADDING_WORDS_NOINIT(0xC);
+
+                PrimitiveTopologyOverride topology_override;
+
+                INSERT_PADDING_WORDS_NOINIT(0x12);
 
                 u32 depth_bounds_enable;
 
@@ -1531,6 +1556,9 @@ private:
     /// Handles a write to the VERTEX_END_GL register, triggering a draw.
     void DrawArrays();
 
+    /// Handles use of topology overrides (e.g., to avoid using a topology assigned from a macro)
+    void ProcessTopologyOverride();
+
     // Handles a instance drawcall from MME
     void StepInstance(MMEDrawMode expected_mode, u32 count);
 
@@ -1569,6 +1597,7 @@ private:
     Upload::State upload_state;
 
     bool execute_on{true};
+    bool use_topology_override{false};
 };
 
 #define ASSERT_REG_POSITION(field_name, position)                                                  \
@@ -1685,6 +1714,7 @@ ASSERT_REG_POSITION(draw, 0x585);
 ASSERT_REG_POSITION(primitive_restart, 0x591);
 ASSERT_REG_POSITION(provoking_vertex_last, 0x5A1);
 ASSERT_REG_POSITION(index_array, 0x5F2);
+ASSERT_REG_POSITION(small_index, 0x5F9);
 ASSERT_REG_POSITION(polygon_offset_clamp, 0x61F);
 ASSERT_REG_POSITION(instanced_arrays, 0x620);
 ASSERT_REG_POSITION(vp_point_size, 0x644);
@@ -1694,6 +1724,7 @@ ASSERT_REG_POSITION(cull_face, 0x648);
 ASSERT_REG_POSITION(pixel_center_integer, 0x649);
 ASSERT_REG_POSITION(viewport_transform_enabled, 0x64B);
 ASSERT_REG_POSITION(view_volume_clip_control, 0x64F);
+ASSERT_REG_POSITION(topology_override, 0x65C);
 ASSERT_REG_POSITION(depth_bounds_enable, 0x66F);
 ASSERT_REG_POSITION(logic_op, 0x671);
 ASSERT_REG_POSITION(clear_buffers, 0x674);

src/video_core/host_shaders/CMakeLists.txt

@@ -14,6 +14,7 @@ set(SHADER_FILES
     convert_d24s8_to_abgr8.frag
     convert_depth_to_float.frag
     convert_float_to_depth.frag
+    convert_s8d24_to_abgr8.frag
     full_screen_triangle.vert
     fxaa.frag
     fxaa.vert

src/video_core/host_shaders/convert_s8d24_to_abgr8.frag

@@ -0,0 +1,23 @@
+// Copyright 2022 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#version 450
+
+layout(binding = 0) uniform sampler2D depth_tex;
+layout(binding = 1) uniform isampler2D stencil_tex;
+
+layout(location = 0) out vec4 color;
+
+void main() {
+    ivec2 coord = ivec2(gl_FragCoord.xy);
+    uint depth = uint(textureLod(depth_tex, coord, 0).r * (exp2(24.0) - 1.0f));
+    uint stencil = uint(textureLod(stencil_tex, coord, 0).r);
+
+    highp uint depth_val =
+        uint(textureLod(depth_tex, coord, 0).r * (exp2(32.0) - 1.0));
+    lowp uint stencil_val = textureLod(stencil_tex, coord, 0).r;
+    highp uvec4 components =
+        uvec4((uvec3(depth_val) >> uvec3(24u, 16u, 8u)) & 0x000000FFu, stencil_val);
+    color.rgba = vec4(components) / (exp2(8.0) - 1.0);
+}

src/video_core/renderer_vulkan/blit_image.cpp

@@ -9,6 +9,7 @@
 #include "video_core/host_shaders/convert_d24s8_to_abgr8_frag_spv.h"
 #include "video_core/host_shaders/convert_depth_to_float_frag_spv.h"
 #include "video_core/host_shaders/convert_float_to_depth_frag_spv.h"
+#include "video_core/host_shaders/convert_s8d24_to_abgr8_frag_spv.h"
 #include "video_core/host_shaders/full_screen_triangle_vert_spv.h"
 #include "video_core/host_shaders/vulkan_blit_color_float_frag_spv.h"
 #include "video_core/host_shaders/vulkan_blit_depth_stencil_frag_spv.h"
@@ -370,6 +371,7 @@ BlitImageHelper::BlitImageHelper(const Device& device_, VKScheduler& scheduler_,
       convert_float_to_depth_frag(BuildShader(device, CONVERT_FLOAT_TO_DEPTH_FRAG_SPV)),
       convert_abgr8_to_d24s8_frag(BuildShader(device, CONVERT_ABGR8_TO_D24S8_FRAG_SPV)),
       convert_d24s8_to_abgr8_frag(BuildShader(device, CONVERT_D24S8_TO_ABGR8_FRAG_SPV)),
+      convert_s8d24_to_abgr8_frag(BuildShader(device, CONVERT_S8D24_TO_ABGR8_FRAG_SPV)),
       linear_sampler(device.GetLogical().CreateSampler(SAMPLER_CREATE_INFO<VK_FILTER_LINEAR>)),
       nearest_sampler(device.GetLogical().CreateSampler(SAMPLER_CREATE_INFO<VK_FILTER_NEAREST>)) {
     if (device.IsExtShaderStencilExportSupported()) {
@@ -474,6 +476,13 @@ void BlitImageHelper::ConvertD24S8ToABGR8(const Framebuffer* dst_framebuffer,
     ConvertDepthStencil(*convert_d24s8_to_abgr8_pipeline, dst_framebuffer, src_image_view);
 }
 
+void BlitImageHelper::ConvertS8D24ToABGR8(const Framebuffer* dst_framebuffer,
+                                          ImageView& src_image_view) {
+    ConvertPipelineColorTargetEx(convert_s8d24_to_abgr8_pipeline, dst_framebuffer->RenderPass(),
+                                 convert_s8d24_to_abgr8_frag);
+    ConvertDepthStencil(*convert_s8d24_to_abgr8_pipeline, dst_framebuffer, src_image_view);
+}
+
 void BlitImageHelper::Convert(VkPipeline pipeline, const Framebuffer* dst_framebuffer,
                               const ImageView& src_image_view) {
     const VkPipelineLayout layout = *one_texture_pipeline_layout;

src/video_core/renderer_vulkan/blit_image.h

@@ -56,6 +56,8 @@ public:
 
     void ConvertD24S8ToABGR8(const Framebuffer* dst_framebuffer, ImageView& src_image_view);
 
+    void ConvertS8D24ToABGR8(const Framebuffer* dst_framebuffer, ImageView& src_image_view);
+
 private:
     void Convert(VkPipeline pipeline, const Framebuffer* dst_framebuffer,
                  const ImageView& src_image_view);
@@ -99,6 +101,7 @@ private:
     vk::ShaderModule convert_float_to_depth_frag;
     vk::ShaderModule convert_abgr8_to_d24s8_frag;
     vk::ShaderModule convert_d24s8_to_abgr8_frag;
+    vk::ShaderModule convert_s8d24_to_abgr8_frag;
     vk::Sampler linear_sampler;
     vk::Sampler nearest_sampler;
 
@@ -112,6 +115,7 @@ private:
     vk::Pipeline convert_r16_to_d16_pipeline;
     vk::Pipeline convert_abgr8_to_d24s8_pipeline;
     vk::Pipeline convert_d24s8_to_abgr8_pipeline;
+    vk::Pipeline convert_s8d24_to_abgr8_pipeline;
 };
 
 } // namespace Vulkan

src/video_core/renderer_vulkan/vk_compute_pass.cpp

@@ -2,6 +2,7 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include <array>
 #include <cstring>
 #include <memory>
 #include <optional>
@@ -292,7 +293,7 @@ std::pair<VkBuffer, VkDeviceSize> QuadIndexedPass::Assemble(
             .srcAccessMask = VK_ACCESS_SHADER_WRITE_BIT,
             .dstAccessMask = VK_ACCESS_INDEX_READ_BIT,
         };
-        const std::array push_constants{base_vertex, index_shift};
+        const std::array<u32, 2> push_constants{base_vertex, index_shift};
         const VkDescriptorSet set = descriptor_allocator.Commit();
         device.GetLogical().UpdateDescriptorSet(set, *descriptor_template, descriptor_data);
         cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_COMPUTE, *pipeline);

src/video_core/renderer_vulkan/vk_texture_cache.cpp

@@ -1070,6 +1070,9 @@ void TextureCacheRuntime::ConvertImage(Framebuffer* dst, ImageView& dst_view, Im
         if (src_view.format == PixelFormat::S8_UINT_D24_UNORM) {
             return blit_image_helper.ConvertD24S8ToABGR8(dst, src_view);
         }
+        if (src_view.format == PixelFormat::D24_UNORM_S8_UINT) {
+            return blit_image_helper.ConvertS8D24ToABGR8(dst, src_view);
+        }
         break;
     case PixelFormat::R32_FLOAT:
         if (src_view.format == PixelFormat::D32_FLOAT) {

src/yuzu/configuration/config.cpp

@@ -1155,6 +1155,8 @@ void Config::SaveCpuValues() {
         WriteBasicSetting(Settings::values.cpuopt_misc_ir);
         WriteBasicSetting(Settings::values.cpuopt_reduce_misalign_checks);
         WriteBasicSetting(Settings::values.cpuopt_fastmem);
+        WriteBasicSetting(Settings::values.cpuopt_fastmem_exclusives);
+        WriteBasicSetting(Settings::values.cpuopt_recompile_exclusives);
     }
 
     qt_config->endGroup();